Files
RLCR-v4-ks-uniqueness-cov0-…/trainer_state.json
ModelHub XC 27aabf08b4 初始化项目,由ModelHub XC社区提供模型
Model: hector-gr/RLCR-v4-ks-uniqueness-cov0-entropy100-noece-noaurc-scaletrue-highcov-batchaccgated-hotpot
Source: Original Platform
2026-05-09 19:55:48 +08:00

9805 lines
629 KiB
JSON

{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.9984,
"eval_steps": 50,
"global_step": 312,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"calibration/aurc": 0.6247499988531867,
"calibration/batch_distribution_entropy": 0.6498143872002083,
"calibration/batch_entropy_100bins": 0.4813110906504229,
"calibration/batch_entropy_10bins": 0.6498143872002083,
"calibration/batch_entropy_50bins": 0.5642144676174681,
"calibration/batch_uniqueness": 0.7239538890297407,
"calibration/confidence_entropy": 0.34517574402231066,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.0,
"calibration/coverage@20%": 0.0,
"calibration/coverage@25%": 0.0,
"calibration/coverage@30%": 0.0,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.6498143872002083,
"calibration/distribution_entropy_100": 0.4813110906504229,
"calibration/ece": 0.4909404991600537,
"calibration/mean_confidence": 0.7900560061227713,
"calibration/unique_confidence_per_question": 0.034765625,
"calibration/unique_confidences": 17.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0349609375,
"completions/max_length": 1497.2,
"completions/max_terminated_length": 1497.2,
"completions/mean_length": 213.99521484375,
"completions/mean_terminated_length": 221.74649963378906,
"completions/min_length": 0.0,
"completions/min_terminated_length": 2.0,
"epoch": 0.016,
"grad_norm": 0.09277201443910599,
"learning_rate": 3.1249999999999997e-07,
"loss": 0.0309,
"num_tokens": 17035343.0,
"reward": 0.49630903005599974,
"reward_std": 0.36898610591888426,
"rewards/accuracy_reward": 0.22294921875,
"rewards/batchaccgated_coverage_0": 0.007798153860494495,
"rewards/batchaccgated_coverage_1": 0.007798153860494495,
"rewards/batchaccgated_coverage_10": 0.03437841534614563,
"rewards/batchaccgated_coverage_15": 0.04447134137153626,
"rewards/batchaccgated_coverage_20": 0.04897649548947811,
"rewards/batchaccgated_coverage_25": 0.05406920611858368,
"rewards/batchaccgated_coverage_5": 0.02111861053854227,
"rewards/brier_reward": 0.37785094380378725,
"rewards/confidence_uniqueness_reward": 0.48845864534378053,
"rewards/format_reward": 0.6828125,
"rewards/frontier_entropy_batch_reward": -0.6506382346153259,
"signal/accuracy_reward/centered_abs_mean": 0.243292236328125,
"signal/accuracy_reward/group_std_mean": 0.28762544989585875,
"signal/accuracy_reward/group_zero_std_frac": 0.3,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.3655422806739807,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.1216461181640625,
"signal/advantage_abs_mean": 0.8426231741905212,
"signal/advantage_pre_scale_abs_mean": 0.31455646753311156,
"signal/advantage_pre_scale_std": 0.38326575756073,
"signal/advantage_std": 0.9841922283172607,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.011150767467916012,
"signal/batchaccgated_coverage_0/group_std_mean": 0.0186555091291666,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.0033660505432635546,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0011150768143124879,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.011150767467916012,
"signal/batchaccgated_coverage_1/group_std_mean": 0.0186555091291666,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.0033660505432635546,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0011150768143124879,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03064337410032749,
"signal/batchaccgated_coverage_10/group_std_mean": 0.040476764738559726,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.009092811122536659,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0030643373960629106,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.04068656116724014,
"signal/batchaccgated_coverage_15/group_std_mean": 0.051607031375169754,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.012192328460514545,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004068656219169497,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04656565524637699,
"signal/batchaccgated_coverage_20/group_std_mean": 0.05778969749808312,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.014000780694186688,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004656565329059959,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.05417446196079254,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06630120053887367,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.016334007680416106,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.005417445953935385,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.019522538781166075,
"signal/batchaccgated_coverage_5/group_std_mean": 0.028476596996188163,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.005803767265751958,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0019522539805620909,
"signal/brier_reward/centered_abs_mean": 0.32002233862876894,
"signal/brier_reward/group_std_mean": 0.3658666670322418,
"signal/brier_reward/group_zero_std_frac": 0.003125,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.09691611230373383,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.03200223371386528,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.2968505024909973,
"signal/confidence_uniqueness_reward/group_std_mean": 0.34797369241714476,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.09025935530662536,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.02968505024909973,
"signal/format_reward/centered_abs_mean": 0.40240478515625,
"signal/format_reward/group_std_mean": 0.4528887033462524,
"signal/format_reward/group_zero_std_frac": 0.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.6113224267959595,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.201202392578125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.42713152170181273,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.47397165894508364,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.12979226559400558,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.04271315336227417,
"step": 5
},
{
"calibration/aurc": 0.6687282184403192,
"calibration/batch_distribution_entropy": 0.6476371312962124,
"calibration/batch_entropy_100bins": 0.4842587177409682,
"calibration/batch_entropy_10bins": 0.6476371312962124,
"calibration/batch_entropy_50bins": 0.5638866213478619,
"calibration/batch_uniqueness": 0.7211674710611267,
"calibration/confidence_entropy": 0.34324053350273015,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.0,
"calibration/coverage@20%": 0.0,
"calibration/coverage@25%": 0.0,
"calibration/coverage@30%": 0.0,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.6476371312962124,
"calibration/distribution_entropy_100": 0.4842587177409682,
"calibration/ece": 0.5256624822988435,
"calibration/mean_confidence": 0.7920013810666202,
"calibration/unique_confidence_per_question": 0.037109375,
"calibration/unique_confidences": 19.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.03515625,
"completions/max_length": 1512.6,
"completions/max_terminated_length": 1512.6,
"completions/mean_length": 199.00537109375,
"completions/mean_terminated_length": 206.30809326171874,
"completions/min_length": 0.0,
"completions/min_terminated_length": 3.0,
"epoch": 0.032,
"grad_norm": 0.03582754358649254,
"learning_rate": 6.249999999999999e-07,
"loss": 0.0182,
"num_tokens": 34173510.0,
"reward": 0.5258957743644714,
"reward_std": 0.3390905916690826,
"rewards/accuracy_reward": 0.21982421875,
"rewards/batchaccgated_coverage_0": 0.014002935215830802,
"rewards/batchaccgated_coverage_1": 0.014002935215830802,
"rewards/batchaccgated_coverage_10": 0.02829661816358566,
"rewards/batchaccgated_coverage_15": 0.037445138394832614,
"rewards/batchaccgated_coverage_20": 0.045209021866321565,
"rewards/batchaccgated_coverage_25": 0.04908238872885704,
"rewards/batchaccgated_coverage_5": 0.0210996238514781,
"rewards/brier_reward": 0.3930462062358856,
"rewards/confidence_uniqueness_reward": 0.5304723620414734,
"rewards/format_reward": 0.7470703125,
"rewards/frontier_entropy_batch_reward": -0.7081723213195801,
"signal/accuracy_reward/centered_abs_mean": 0.227532958984375,
"signal/accuracy_reward/group_std_mean": 0.27006119191646577,
"signal/accuracy_reward/group_zero_std_frac": 0.3375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.37521023750305177,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.1137664794921875,
"signal/advantage_abs_mean": 0.8083276271820068,
"signal/advantage_pre_scale_abs_mean": 0.2793608784675598,
"signal/advantage_pre_scale_std": 0.35301142930984497,
"signal/advantage_std": 0.9841665148735046,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.01256183795630932,
"signal/batchaccgated_coverage_0/group_std_mean": 0.020442500337958337,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.0041111950296908615,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.001256183790974319,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.01256183795630932,
"signal/batchaccgated_coverage_1/group_std_mean": 0.020442500337958337,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.0041111950296908615,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.001256183790974319,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.021723009273409845,
"signal/batchaccgated_coverage_10/group_std_mean": 0.03058185279369354,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.007120975758880377,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0021723009180277587,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.029889940470457076,
"signal/batchaccgated_coverage_15/group_std_mean": 0.039534883573651314,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.009805875085294246,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.002988994074985385,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.03890311121940613,
"signal/batchaccgated_coverage_20/group_std_mean": 0.04982725381851196,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.012776435166597367,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.003890311298891902,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.044103769958019255,
"signal/batchaccgated_coverage_25/group_std_mean": 0.0556812547147274,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.0145026296377182,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.0044103770982474085,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.016086464747786522,
"signal/batchaccgated_coverage_5/group_std_mean": 0.025005444884300232,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.0053304544184356924,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0016086465446278452,
"signal/brier_reward/centered_abs_mean": 0.3066942811012268,
"signal/brier_reward/group_std_mean": 0.354111921787262,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1011395812034607,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.030669428408145905,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.26974273324012754,
"signal/confidence_uniqueness_reward/group_std_mean": 0.33063756823539736,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.08872608542442321,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.026974274218082427,
"signal/format_reward/centered_abs_mean": 0.35228271484375,
"signal/format_reward/group_std_mean": 0.4218889534473419,
"signal/format_reward/group_zero_std_frac": 0.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.5779661178588867,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.176141357421875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.3877986788749695,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.4501940131187439,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.12743044048547744,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.038779867440462114,
"step": 10
},
{
"calibration/aurc": 0.5910455432596686,
"calibration/batch_distribution_entropy": 0.6307977471053288,
"calibration/batch_entropy_100bins": 0.47885485207861106,
"calibration/batch_entropy_10bins": 0.6307977471053288,
"calibration/batch_entropy_50bins": 0.5578180522125267,
"calibration/batch_uniqueness": 0.7039643516804514,
"calibration/buffer_distribution_entropy": 0.654344010892302,
"calibration/buffer_entropy_100bins": 0.4900749095121335,
"calibration/buffer_entropy_10bins": 0.654344010892302,
"calibration/buffer_entropy_50bins": 0.5718915776286453,
"calibration/confidence_entropy": 0.3331853446170793,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.0,
"calibration/coverage@20%": 0.0,
"calibration/coverage@25%": 0.0,
"calibration/coverage@30%": 0.0,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.6307977471053288,
"calibration/distribution_entropy_100": 0.47885485207861106,
"calibration/ece": 0.476347096177906,
"calibration/mean_confidence": 0.8094632548714371,
"calibration/unique_confidence_per_question": 0.039453125,
"calibration/unique_confidences": 20.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.01396484375,
"completions/max_length": 1307.0,
"completions/max_terminated_length": 1307.0,
"completions/mean_length": 163.44501953125,
"completions/mean_terminated_length": 165.8753234863281,
"completions/min_length": 0.0,
"completions/min_terminated_length": 19.0,
"epoch": 0.048,
"grad_norm": 0.034932106733322144,
"learning_rate": 9.374999999999999e-07,
"loss": 0.0144,
"num_tokens": 50895923.0,
"reward": 0.662801194190979,
"reward_std": 0.2578377366065979,
"rewards/accuracy_reward": 0.2810546875,
"rewards/batchaccgated_coverage_0": 0.01390560856088996,
"rewards/batchaccgated_coverage_1": 0.01390560856088996,
"rewards/batchaccgated_coverage_10": 0.06534868441522121,
"rewards/batchaccgated_coverage_15": 0.07830065339803696,
"rewards/batchaccgated_coverage_20": 0.08644865602254867,
"rewards/batchaccgated_coverage_25": 0.0945888377726078,
"rewards/batchaccgated_coverage_5": 0.03706094361841679,
"rewards/brier_reward": 0.49367535710334776,
"rewards/confidence_uniqueness_reward": 0.6600236296653748,
"rewards/format_reward": 0.90751953125,
"rewards/frontier_entropy_batch_reward": -0.8581171035766602,
"signal/accuracy_reward/centered_abs_mean": 0.189453125,
"signal/accuracy_reward/group_std_mean": 0.23639306724071502,
"signal/accuracy_reward/group_zero_std_frac": 0.384375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.4933402955532074,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0947265625,
"signal/advantage_abs_mean": 0.6934794306755065,
"signal/advantage_pre_scale_abs_mean": 0.18734275698661804,
"signal/advantage_pre_scale_std": 0.2764400362968445,
"signal/advantage_std": 0.9839721918106079,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.013470524176955222,
"signal/batchaccgated_coverage_0/group_std_mean": 0.023414578661322592,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.38125,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.007230181712657213,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0013470524689182639,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.013470524176955222,
"signal/batchaccgated_coverage_1/group_std_mean": 0.023414578661322592,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.38125,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.007230181712657213,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0013470524689182639,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03313998244702816,
"signal/batchaccgated_coverage_10/group_std_mean": 0.04966066852211952,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.38125,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.01879340410232544,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0033139981562271712,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.04224871546030044,
"signal/batchaccgated_coverage_15/group_std_mean": 0.060867539048194884,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.38125,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.02294488623738289,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004224871471524239,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.049050360918045044,
"signal/batchaccgated_coverage_20/group_std_mean": 0.0685936838388443,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.38125,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.02593726236373186,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004905036184936762,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.05957357734441757,
"signal/batchaccgated_coverage_25/group_std_mean": 0.08025077879428863,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.38125,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.031515516340732574,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.0059573578648269175,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.020516308769583703,
"signal/batchaccgated_coverage_5/group_std_mean": 0.03220605403184891,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.38125,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.011259878054261208,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0020516308955848216,
"signal/brier_reward/centered_abs_mean": 0.2624696552753448,
"signal/brier_reward/group_std_mean": 0.3185483515262604,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.13646066188812256,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.026246964931488037,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.18760752975940703,
"signal/confidence_uniqueness_reward/group_std_mean": 0.24785077273845674,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.09555385410785674,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.01876075379550457,
"signal/format_reward/centered_abs_mean": 0.157489013671875,
"signal/format_reward/group_std_mean": 0.258052384853363,
"signal/format_reward/group_zero_std_frac": 0.1125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.36937561333179475,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0787445068359375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.23125048577785492,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3432617664337158,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.03125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.11669281423091889,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.023125049099326132,
"step": 15
},
{
"calibration/aurc": 0.530962046220613,
"calibration/batch_distribution_entropy": 0.7386738988189596,
"calibration/batch_entropy_100bins": 0.5636245805953478,
"calibration/batch_entropy_10bins": 0.7386738988189596,
"calibration/batch_entropy_50bins": 0.6436320596408155,
"calibration/batch_uniqueness": 0.7939807262869509,
"calibration/buffer_distribution_entropy": 0.6573260567307225,
"calibration/buffer_entropy_100bins": 0.49729805165130603,
"calibration/buffer_entropy_10bins": 0.6573260567307225,
"calibration/buffer_entropy_50bins": 0.5781749667962972,
"calibration/confidence_entropy": 0.3878956759898032,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.0,
"calibration/coverage@20%": 0.0,
"calibration/coverage@25%": 0.0,
"calibration/coverage@30%": 0.0,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.7386738988189596,
"calibration/distribution_entropy_100": 0.5636245805953478,
"calibration/ece": 0.3514334095946649,
"calibration/mean_confidence": 0.7551914150233393,
"calibration/unique_confidence_per_question": 0.076953125,
"calibration/unique_confidences": 39.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00283203125,
"completions/max_length": 946.0,
"completions/max_terminated_length": 946.0,
"completions/mean_length": 124.1779296875,
"completions/mean_terminated_length": 124.53269805908204,
"completions/min_length": 0.0,
"completions/min_terminated_length": 30.8,
"epoch": 0.064,
"grad_norm": 0.04307975247502327,
"learning_rate": 1e-06,
"loss": -0.0115,
"num_tokens": 67085905.0,
"reward": 0.7622903227806092,
"reward_std": 0.20451631844043733,
"rewards/accuracy_reward": 0.35380859375,
"rewards/batchaccgated_coverage_0": 0.00871854261495173,
"rewards/batchaccgated_coverage_1": 0.00871854261495173,
"rewards/batchaccgated_coverage_10": 0.0670653223991394,
"rewards/batchaccgated_coverage_15": 0.08054911568760872,
"rewards/batchaccgated_coverage_20": 0.09976833462715148,
"rewards/batchaccgated_coverage_25": 0.10426986515522003,
"rewards/batchaccgated_coverage_5": 0.029751383885741233,
"rewards/brier_reward": 0.5949586033821106,
"rewards/confidence_uniqueness_reward": 0.7916639924049378,
"rewards/format_reward": 0.98515625,
"rewards/frontier_entropy_batch_reward": -0.85738445520401,
"signal/accuracy_reward/centered_abs_mean": 0.204986572265625,
"signal/accuracy_reward/group_std_mean": 0.25679649114608766,
"signal/accuracy_reward/group_zero_std_frac": 0.325,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7098461627960205,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.1024932861328125,
"signal/advantage_abs_mean": 0.722055447101593,
"signal/advantage_pre_scale_abs_mean": 0.15270549952983856,
"signal/advantage_pre_scale_std": 0.2225021243095398,
"signal/advantage_std": 0.9837960124015808,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.022280838713049887,
"signal/batchaccgated_coverage_0/group_std_mean": 0.036336852610111235,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.016128088161349297,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0022280839271843433,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.022280838713049887,
"signal/batchaccgated_coverage_1/group_std_mean": 0.036336852610111235,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.016128088161349297,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0022280839271843433,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.0419327475130558,
"signal/batchaccgated_coverage_10/group_std_mean": 0.06157161220908165,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.029415056109428406,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0041932746767997745,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.049335777014493945,
"signal/batchaccgated_coverage_15/group_std_mean": 0.07056339979171752,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.03478674814105034,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0049335777759552,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.06370371505618096,
"signal/batchaccgated_coverage_20/group_std_mean": 0.0885491669178009,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.04435234814882279,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.006370371580123902,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.06832471266388893,
"signal/batchaccgated_coverage_25/group_std_mean": 0.09369181990623474,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.047871388494968414,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.006832471396774053,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.027172042429447173,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04238997101783752,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.019730507768690585,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.002717204345390201,
"signal/brier_reward/centered_abs_mean": 0.2422477751970291,
"signal/brier_reward/group_std_mean": 0.29777374267578127,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.17120460867881776,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.024224777519702912,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.09979309886693954,
"signal/confidence_uniqueness_reward/group_std_mean": 0.1357775703072548,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.06854969114065171,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.009979309700429439,
"signal/format_reward/centered_abs_mean": 0.02833251953125,
"signal/format_reward/group_std_mean": 0.0733065165579319,
"signal/format_reward/group_zero_std_frac": 0.621875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.09608695805072784,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.014166259765625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2355801373720169,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3643790125846863,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.05,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.17129166871309282,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.023558012768626214,
"step": 20
},
{
"calibration/aurc": 0.6637611945793297,
"calibration/batch_distribution_entropy": 0.9606030171869822,
"calibration/batch_entropy_100bins": 0.8820346311430585,
"calibration/batch_entropy_10bins": 0.9606030171869822,
"calibration/batch_entropy_50bins": 0.923518711059209,
"calibration/batch_uniqueness": 0.9418444130813276,
"calibration/buffer_distribution_entropy": 0.7292711046134217,
"calibration/buffer_entropy_100bins": 0.5760054475825083,
"calibration/buffer_entropy_10bins": 0.7292711046134217,
"calibration/buffer_entropy_50bins": 0.6540098785189122,
"calibration/confidence_entropy": 0.4880019724332752,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.0,
"calibration/coverage@20%": 0.0,
"calibration/coverage@25%": 0.0,
"calibration/coverage@30%": 0.0,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.9606030171869822,
"calibration/distribution_entropy_100": 0.8820346311430585,
"calibration/ece": 0.32463927062643827,
"calibration/mean_confidence": 0.5739366408271904,
"calibration/unique_confidence_per_question": 0.38203125,
"calibration/unique_confidences": 195.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0017578125,
"completions/max_length": 745.2,
"completions/max_terminated_length": 745.2,
"completions/mean_length": 105.29404296875,
"completions/mean_terminated_length": 105.47904663085937,
"completions/min_length": 0.0,
"completions/min_terminated_length": 38.2,
"epoch": 0.08,
"grad_norm": 0.08280298858880997,
"learning_rate": 1e-06,
"loss": -0.0113,
"num_tokens": 83097268.0,
"reward": 0.8279740691184998,
"reward_std": 0.17523195445537568,
"rewards/accuracy_reward": 0.3578125,
"rewards/batchaccgated_coverage_0": 0.010622684704139828,
"rewards/batchaccgated_coverage_1": 0.010622684704139828,
"rewards/batchaccgated_coverage_10": 0.05529484562575817,
"rewards/batchaccgated_coverage_15": 0.06275972127914428,
"rewards/batchaccgated_coverage_20": 0.0706663429737091,
"rewards/batchaccgated_coverage_25": 0.0762055642902851,
"rewards/batchaccgated_coverage_5": 0.035876252315938474,
"rewards/brier_reward": 0.6663513898849487,
"rewards/confidence_uniqueness_reward": 0.937761652469635,
"rewards/format_reward": 0.9953125,
"rewards/frontier_entropy_batch_reward": -0.4120453268289566,
"signal/accuracy_reward/centered_abs_mean": 0.19283447265625,
"signal/accuracy_reward/group_std_mean": 0.24174462556838988,
"signal/accuracy_reward/group_zero_std_frac": 0.35625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.77999347448349,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.096417236328125,
"signal/advantage_abs_mean": 0.7750329494476318,
"signal/advantage_pre_scale_abs_mean": 0.13666608780622483,
"signal/advantage_pre_scale_std": 0.19091297686100006,
"signal/advantage_std": 0.9836948990821839,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.05143122002482414,
"signal/batchaccgated_coverage_0/group_std_mean": 0.06799714267253876,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.04193255566060543,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.005143122002482414,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.05143122002482414,
"signal/batchaccgated_coverage_1/group_std_mean": 0.06799714267253876,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.04193255566060543,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.005143122002482414,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.06415941789746285,
"signal/batchaccgated_coverage_10/group_std_mean": 0.0823859766125679,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.052013760805130003,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.006415942125022412,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.06605891212821006,
"signal/batchaccgated_coverage_15/group_std_mean": 0.08451238870620728,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05356289893388748,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0066058912314474584,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.06851867213845253,
"signal/batchaccgated_coverage_20/group_std_mean": 0.08735992908477783,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.0553885467350483,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0068518673069775105,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.07287217453122138,
"signal/batchaccgated_coverage_25/group_std_mean": 0.09267176836729049,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.058624989539384845,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.007287217397242784,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.05820987448096275,
"signal/batchaccgated_coverage_5/group_std_mean": 0.07511185705661774,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.04736398532986641,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.005820987466722727,
"signal/brier_reward/centered_abs_mean": 0.24181779325008393,
"signal/brier_reward/group_std_mean": 0.29060696959495547,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.19556585252285003,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.024181779474020004,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.033458396419882776,
"signal/confidence_uniqueness_reward/group_std_mean": 0.05254848524928093,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.026074886322021484,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.003345839632675052,
"signal/format_reward/centered_abs_mean": 0.009033203125,
"signal/format_reward/group_std_mean": 0.025171270594000816,
"signal/format_reward/group_zero_std_frac": 0.8625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0359637625515461,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0045166015625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.4187507629394531,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.4902602732181549,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.33579265475273135,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.04187507778406143,
"step": 25
},
{
"calibration/aurc": 0.6446439673215554,
"calibration/batch_distribution_entropy": 0.9495266461127774,
"calibration/batch_entropy_100bins": 0.9473326059710707,
"calibration/batch_entropy_10bins": 0.9495266461127774,
"calibration/batch_entropy_50bins": 0.9574622044480611,
"calibration/batch_uniqueness": 0.9479758820065257,
"calibration/buffer_distribution_entropy": 0.831525997594637,
"calibration/buffer_entropy_100bins": 0.7130262014148991,
"calibration/buffer_entropy_10bins": 0.831525997594637,
"calibration/buffer_entropy_50bins": 0.7764620966499411,
"calibration/confidence_entropy": 0.505928554358938,
"calibration/coverage@0%": 0.003551904093994502,
"calibration/coverage@1%": 0.003551904093994502,
"calibration/coverage@10%": 0.003551904093994502,
"calibration/coverage@15%": 0.003551904093994502,
"calibration/coverage@20%": 0.00473689833306338,
"calibration/coverage@25%": 0.005521212058553576,
"calibration/coverage@30%": 0.006310166693662057,
"calibration/coverage@5%": 0.003551904093994502,
"calibration/distribution_entropy_10": 0.9495266461127774,
"calibration/distribution_entropy_100": 0.9473326059710707,
"calibration/ece": 0.21381107003926095,
"calibration/mean_confidence": 0.3837111766497221,
"calibration/unique_confidence_per_question": 0.864453125,
"calibration/unique_confidences": 442.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0041015625,
"completions/max_length": 768.2,
"completions/max_terminated_length": 768.2,
"completions/mean_length": 112.135546875,
"completions/mean_terminated_length": 112.60049438476562,
"completions/min_length": 0.0,
"completions/min_terminated_length": 41.8,
"epoch": 0.096,
"grad_norm": 0.024549348279833794,
"learning_rate": 1e-06,
"loss": -0.0122,
"num_tokens": 99290144.0,
"reward": 0.8326921105384827,
"reward_std": 0.15945081710815429,
"rewards/accuracy_reward": 0.36435546875,
"rewards/batchaccgated_coverage_0": -0.0015749878250062466,
"rewards/batchaccgated_coverage_1": -0.0015749878250062466,
"rewards/batchaccgated_coverage_10": 0.03344577793031931,
"rewards/batchaccgated_coverage_15": 0.0404186025261879,
"rewards/batchaccgated_coverage_20": 0.045605845749378204,
"rewards/batchaccgated_coverage_25": 0.04895867630839348,
"rewards/batchaccgated_coverage_5": 0.015098475478589534,
"rewards/brier_reward": 0.714118480682373,
"rewards/confidence_uniqueness_reward": 0.9427257657051087,
"rewards/format_reward": 0.99345703125,
"rewards/frontier_entropy_batch_reward": -0.2993633210659027,
"signal/accuracy_reward/centered_abs_mean": 0.182537841796875,
"signal/accuracy_reward/group_std_mean": 0.23251287043094634,
"signal/accuracy_reward/group_zero_std_frac": 0.371875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9863747835159302,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0912689208984375,
"signal/advantage_abs_mean": 0.7647687792778015,
"signal/advantage_pre_scale_abs_mean": 0.12076428234577179,
"signal/advantage_pre_scale_std": 0.1793455958366394,
"signal/advantage_std": 0.9834189295768738,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.055737099051475524,
"signal/batchaccgated_coverage_0/group_std_mean": 0.07064625471830369,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.33125,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06005018278956413,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.005573710147291422,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.055737099051475524,
"signal/batchaccgated_coverage_1/group_std_mean": 0.07064625471830369,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.33125,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06005018278956413,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.005573710147291422,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.062380362302064896,
"signal/batchaccgated_coverage_10/group_std_mean": 0.07934080809354782,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.33125,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06708819717168808,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.006238036323338747,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.06268596053123474,
"signal/batchaccgated_coverage_15/group_std_mean": 0.08017706871032715,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.33125,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06733292415738105,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.006268596276640892,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.06353392526507377,
"signal/batchaccgated_coverage_20/group_std_mean": 0.0816449448466301,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.33125,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.06819377839565277,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.006353392824530601,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.06326589584350586,
"signal/batchaccgated_coverage_25/group_std_mean": 0.08189980164170266,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.33125,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.06789372563362121,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.006326589640229941,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.06027383357286453,
"signal/batchaccgated_coverage_5/group_std_mean": 0.07603656202554702,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.33125,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.0648527555167675,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.006027383636683226,
"signal/brier_reward/centered_abs_mean": 0.21727054417133332,
"signal/brier_reward/group_std_mean": 0.2678825318813324,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.23644972145557402,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.021727054566144943,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.02690122611820698,
"signal/confidence_uniqueness_reward/group_std_mean": 0.04977017119526863,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.029332310333847998,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.002690122649073601,
"signal/format_reward/centered_abs_mean": 0.012530517578125,
"signal/format_reward/group_std_mean": 0.03325401395559311,
"signal/format_reward/group_zero_std_frac": 0.825,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.06870068311691284,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0062652587890625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.3750212550163269,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.4427333354949951,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.40717302560806273,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.03750212714076042,
"step": 30
},
{
"calibration/aurc": 0.49302644455978506,
"calibration/batch_distribution_entropy": 0.9815448321519529,
"calibration/batch_entropy_100bins": 0.969913591453523,
"calibration/batch_entropy_10bins": 0.9815448321519529,
"calibration/batch_entropy_50bins": 0.9767011648059878,
"calibration/batch_uniqueness": 0.9537243785144522,
"calibration/buffer_distribution_entropy": 0.8939313618688894,
"calibration/buffer_entropy_100bins": 0.797125701064717,
"calibration/buffer_entropy_10bins": 0.8939313618688894,
"calibration/buffer_entropy_50bins": 0.8473559247003287,
"calibration/confidence_entropy": 0.512242381797994,
"calibration/coverage@0%": 0.0027512828734904482,
"calibration/coverage@1%": 0.0027512828734904482,
"calibration/coverage@10%": 0.00786939310971092,
"calibration/coverage@15%": 0.009444196259317219,
"calibration/coverage@20%": 0.012200101771128244,
"calibration/coverage@25%": 0.015740321008429464,
"calibration/coverage@30%": 0.04289023609257335,
"calibration/coverage@5%": 0.0027512828734904482,
"calibration/distribution_entropy_10": 0.9815448321519529,
"calibration/distribution_entropy_100": 0.969913591453523,
"calibration/ece": 0.20163770493851335,
"calibration/mean_confidence": 0.5388831143375251,
"calibration/unique_confidence_per_question": 0.987890625,
"calibration/unique_confidences": 505.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.003515625,
"completions/max_length": 1079.2,
"completions/max_terminated_length": 1079.2,
"completions/mean_length": 128.69541015625,
"completions/mean_terminated_length": 129.14971160888672,
"completions/min_length": 0.0,
"completions/min_terminated_length": 51.4,
"epoch": 0.112,
"grad_norm": 0.07550515234470367,
"learning_rate": 1e-06,
"loss": -0.0179,
"num_tokens": 115717457.0,
"reward": 0.9012673258781433,
"reward_std": 0.17178937196731567,
"rewards/accuracy_reward": 0.4228515625,
"rewards/batchaccgated_coverage_0": 0.02760154437273741,
"rewards/batchaccgated_coverage_1": 0.02760154437273741,
"rewards/batchaccgated_coverage_10": 0.0756085492670536,
"rewards/batchaccgated_coverage_15": 0.08606690391898156,
"rewards/batchaccgated_coverage_20": 0.09380115717649459,
"rewards/batchaccgated_coverage_25": 0.09888030588626862,
"rewards/batchaccgated_coverage_5": 0.055540557205677035,
"rewards/brier_reward": 0.6986034870147705,
"rewards/confidence_uniqueness_reward": 0.9492161870002747,
"rewards/format_reward": 0.99521484375,
"rewards/frontier_entropy_batch_reward": -0.1905791014432907,
"signal/accuracy_reward/centered_abs_mean": 0.17777099609375,
"signal/accuracy_reward/group_std_mean": 0.22678249776363374,
"signal/accuracy_reward/group_zero_std_frac": 0.378125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.8302661180496216,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.088885498046875,
"signal/advantage_abs_mean": 0.7588428854942322,
"signal/advantage_pre_scale_abs_mean": 0.1317495048046112,
"signal/advantage_pre_scale_std": 0.1888563632965088,
"signal/advantage_std": 0.9835710763931275,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.06383616700768471,
"signal/batchaccgated_coverage_0/group_std_mean": 0.08184875547885895,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.05918530747294426,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.00638361657038331,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.06383616700768471,
"signal/batchaccgated_coverage_1/group_std_mean": 0.08184875547885895,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.05918530747294426,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.00638361657038331,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.07948446199297905,
"signal/batchaccgated_coverage_10/group_std_mean": 0.1003252848982811,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07319761589169502,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.00794844664633274,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.0813552588224411,
"signal/batchaccgated_coverage_15/group_std_mean": 0.10270120054483414,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.0748669870197773,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.008135525695979594,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.08369389101862908,
"signal/batchaccgated_coverage_20/group_std_mean": 0.10533523857593537,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.07683077305555344,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.008369388990104198,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.08560860380530358,
"signal/batchaccgated_coverage_25/group_std_mean": 0.10748820304870606,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.0782657042145729,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.008560860436409712,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.07434996291995048,
"signal/batchaccgated_coverage_5/group_std_mean": 0.09397305250167846,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06868454366922379,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.007434996217489243,
"signal/brier_reward/centered_abs_mean": 0.2259790390729904,
"signal/brier_reward/group_std_mean": 0.27508466243743895,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.21150386929512024,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.02259790413081646,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01930009312927723,
"signal/confidence_uniqueness_reward/group_std_mean": 0.03837290070950985,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.017997850850224494,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0019300092943012714,
"signal/format_reward/centered_abs_mean": 0.009246826171875,
"signal/format_reward/group_std_mean": 0.026396314054727553,
"signal/format_reward/group_zero_std_frac": 0.853125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.043405921384692195,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0046234130859375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2753020167350769,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3534925222396851,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.2554273694753647,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02753020226955414,
"step": 35
},
{
"calibration/aurc": 0.5512790072041012,
"calibration/batch_distribution_entropy": 0.9387721206346997,
"calibration/batch_entropy_100bins": 0.9485455329411432,
"calibration/batch_entropy_10bins": 0.9387721206346997,
"calibration/batch_entropy_50bins": 0.9530534666580806,
"calibration/batch_uniqueness": 0.9465176729714759,
"calibration/buffer_distribution_entropy": 0.9130218758428812,
"calibration/buffer_entropy_100bins": 0.843914085015126,
"calibration/buffer_entropy_10bins": 0.9130218758428812,
"calibration/buffer_entropy_50bins": 0.8823592401930892,
"calibration/confidence_entropy": 0.48987493681147837,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.0,
"calibration/coverage@20%": 0.00234375,
"calibration/coverage@25%": 0.00390625,
"calibration/coverage@30%": 0.004296875,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.9387721206346997,
"calibration/distribution_entropy_100": 0.9485455329411432,
"calibration/ece": 0.2684756227486903,
"calibration/mean_confidence": 0.6406591087270314,
"calibration/unique_confidence_per_question": 0.995703125,
"calibration/unique_confidences": 509.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0041015625,
"completions/max_length": 995.6,
"completions/max_terminated_length": 995.6,
"completions/mean_length": 141.88388671875,
"completions/mean_terminated_length": 142.4716003417969,
"completions/min_length": 0.0,
"completions/min_terminated_length": 57.0,
"epoch": 0.128,
"grad_norm": 0.013333677314221859,
"learning_rate": 1e-06,
"loss": -0.0198,
"num_tokens": 132087020.0,
"reward": 0.9167538762092591,
"reward_std": 0.17794648110866546,
"rewards/accuracy_reward": 0.4267578125,
"rewards/batchaccgated_coverage_0": 0.05797231309115887,
"rewards/batchaccgated_coverage_1": 0.05797231309115887,
"rewards/batchaccgated_coverage_10": 0.1185494139790535,
"rewards/batchaccgated_coverage_15": 0.1279773786664009,
"rewards/batchaccgated_coverage_20": 0.13704620897769929,
"rewards/batchaccgated_coverage_25": 0.14341241121292114,
"rewards/batchaccgated_coverage_5": 0.09276288896799087,
"rewards/brier_reward": 0.673115861415863,
"rewards/confidence_uniqueness_reward": 0.9428873181343078,
"rewards/format_reward": 0.99541015625,
"rewards/frontier_entropy_batch_reward": -0.2949973285198212,
"signal/accuracy_reward/centered_abs_mean": 0.15020751953125,
"signal/accuracy_reward/group_std_mean": 0.20000874102115632,
"signal/accuracy_reward/group_zero_std_frac": 0.421875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.5678041756153107,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.075103759765625,
"signal/advantage_abs_mean": 0.7641587972640991,
"signal/advantage_pre_scale_abs_mean": 0.13507477939128876,
"signal/advantage_pre_scale_std": 0.19270414113998413,
"signal/advantage_std": 0.9837418913841247,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.059799878299236296,
"signal/batchaccgated_coverage_0/group_std_mean": 0.07858101502060891,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.04515654742717743,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.005979987978935242,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.059799878299236296,
"signal/batchaccgated_coverage_1/group_std_mean": 0.07858101502060891,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.04515654742717743,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.005979987978935242,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.08110829889774322,
"signal/batchaccgated_coverage_10/group_std_mean": 0.1039955273270607,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06173477098345757,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.008110829815268516,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.08263948112726212,
"signal/batchaccgated_coverage_15/group_std_mean": 0.10548398047685623,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06288884282112121,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.00826394846662879,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.08840580135583878,
"signal/batchaccgated_coverage_20/group_std_mean": 0.11262985914945603,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.0671466164290905,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.008840580191463231,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.09443405270576477,
"signal/batchaccgated_coverage_25/group_std_mean": 0.11931047588586807,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.07172784358263015,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.009443405643105507,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.06997200846672058,
"signal/batchaccgated_coverage_5/group_std_mean": 0.09014314413070679,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.053158923983573914,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.006997200567275286,
"signal/brier_reward/centered_abs_mean": 0.2249685287475586,
"signal/brier_reward/group_std_mean": 0.2747365415096283,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1714296370744705,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.0224968533962965,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.024050904810428618,
"signal/confidence_uniqueness_reward/group_std_mean": 0.04204035773873329,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.018337027169764042,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.002405090443789959,
"signal/format_reward/centered_abs_mean": 0.008807373046875,
"signal/format_reward/group_std_mean": 0.023887013643980028,
"signal/format_reward/group_zero_std_frac": 0.871875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.03379032053053379,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0044036865234375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.3622118830680847,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.43490626811981203,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.27538323402404785,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.03622118979692459,
"step": 40
},
{
"calibration/aurc": 0.3062918521976175,
"calibration/batch_distribution_entropy": 0.9615785922592435,
"calibration/batch_entropy_100bins": 0.9591723194331403,
"calibration/batch_entropy_10bins": 0.9615785922592435,
"calibration/batch_entropy_50bins": 0.9648599522289034,
"calibration/batch_uniqueness": 0.9496407942408076,
"calibration/buffer_distribution_entropy": 0.9212058883925511,
"calibration/buffer_entropy_100bins": 0.8729225899703108,
"calibration/buffer_entropy_10bins": 0.9212058883925511,
"calibration/buffer_entropy_50bins": 0.9028848990195286,
"calibration/confidence_entropy": 0.45437588909399046,
"calibration/coverage@0%": 0.007059609087945393,
"calibration/coverage@1%": 0.007059609087945393,
"calibration/coverage@10%": 0.07372627575461206,
"calibration/coverage@15%": 0.18162738722929578,
"calibration/coverage@20%": 0.2361543012200026,
"calibration/coverage@25%": 0.3010354870024809,
"calibration/coverage@30%": 0.5178767893739311,
"calibration/coverage@5%": 0.00980470712716108,
"calibration/distribution_entropy_10": 0.9615785922592435,
"calibration/distribution_entropy_100": 0.9591723194331403,
"calibration/ece": 0.21065046708233207,
"calibration/mean_confidence": 0.5902357243001877,
"calibration/unique_confidence_per_question": 0.9953125,
"calibration/unique_confidences": 509.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00498046875,
"completions/max_length": 1091.2,
"completions/max_terminated_length": 1091.2,
"completions/mean_length": 157.01240234375,
"completions/mean_terminated_length": 157.8075408935547,
"completions/min_length": 0.0,
"completions/min_terminated_length": 62.2,
"epoch": 0.144,
"grad_norm": 0.008763757534325123,
"learning_rate": 1e-06,
"loss": -0.0224,
"num_tokens": 148645259.0,
"reward": 1.0135100722312926,
"reward_std": 0.1873625934123993,
"rewards/accuracy_reward": 0.5341796875,
"rewards/batchaccgated_coverage_0": 0.105792236328125,
"rewards/batchaccgated_coverage_1": 0.105792236328125,
"rewards/batchaccgated_coverage_10": 0.16617338061332704,
"rewards/batchaccgated_coverage_15": 0.17498813569545746,
"rewards/batchaccgated_coverage_20": 0.18244454860687256,
"rewards/batchaccgated_coverage_25": 0.1887723296880722,
"rewards/batchaccgated_coverage_5": 0.13916202485561371,
"rewards/brier_reward": 0.7188340187072754,
"rewards/confidence_uniqueness_reward": 0.946603786945343,
"rewards/format_reward": 0.99482421875,
"rewards/frontier_entropy_batch_reward": -0.23848146796226502,
"signal/accuracy_reward/centered_abs_mean": 0.1519775390625,
"signal/accuracy_reward/group_std_mean": 0.1977204740047455,
"signal/accuracy_reward/group_zero_std_frac": 0.446875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.5570285499095917,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.07598876953125,
"signal/advantage_abs_mean": 0.7696815371513367,
"signal/advantage_pre_scale_abs_mean": 0.14458003640174866,
"signal/advantage_pre_scale_std": 0.20025670528411865,
"signal/advantage_std": 0.9837697267532348,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.09315220415592193,
"signal/batchaccgated_coverage_0/group_std_mean": 0.11760110855102539,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.24375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06807189732789994,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.009315220452845097,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.09315220415592193,
"signal/batchaccgated_coverage_1/group_std_mean": 0.11760110855102539,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.24375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06807189732789994,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.009315220452845097,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.11130553632974624,
"signal/batchaccgated_coverage_10/group_std_mean": 0.13849669247865676,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.24375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08152382969856262,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.011130553856492042,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.114348503947258,
"signal/batchaccgated_coverage_15/group_std_mean": 0.14185929894447327,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.24375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.08377196490764618,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.011434851028025151,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.11737803369760513,
"signal/batchaccgated_coverage_20/group_std_mean": 0.1454198867082596,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.24375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.0857890322804451,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.011737803556025028,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.12140866219997407,
"signal/batchaccgated_coverage_25/group_std_mean": 0.1499514698982239,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.24375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.0886124849319458,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.012140866741538048,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.10452217608690262,
"signal/batchaccgated_coverage_5/group_std_mean": 0.1305285319685936,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.24375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07660740464925767,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.010452217794954777,
"signal/brier_reward/centered_abs_mean": 0.22651352286338805,
"signal/brier_reward/group_std_mean": 0.277605938911438,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.16591331660747527,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.022651352733373643,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.02300041727721691,
"signal/confidence_uniqueness_reward/group_std_mean": 0.042073726654052734,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.016723754443228246,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0023000416811555626,
"signal/format_reward/centered_abs_mean": 0.009930419921875,
"signal/format_reward/group_std_mean": 0.02658817209303379,
"signal/format_reward/group_zero_std_frac": 0.859375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.03550950922071934,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0049652099609375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.33249465227127073,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.41012783646583556,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.24391528964042664,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.03324946314096451,
"step": 45
},
{
"calibration/aurc": 0.3585521676398696,
"calibration/batch_distribution_entropy": 0.9300342218106076,
"calibration/batch_entropy_100bins": 0.9450333391511456,
"calibration/batch_entropy_10bins": 0.9300342218106076,
"calibration/batch_entropy_50bins": 0.9475166855960504,
"calibration/batch_uniqueness": 0.9425474699968355,
"calibration/buffer_distribution_entropy": 0.9280619127547161,
"calibration/buffer_entropy_100bins": 0.8932394343601946,
"calibration/buffer_entropy_10bins": 0.9280619127547161,
"calibration/buffer_entropy_50bins": 0.9172373589463889,
"calibration/confidence_entropy": 0.4271961135067045,
"calibration/coverage@0%": 0.003938597405404571,
"calibration/coverage@1%": 0.003938597405404571,
"calibration/coverage@10%": 0.005122029358067293,
"calibration/coverage@15%": 0.010992870845347136,
"calibration/coverage@20%": 0.052953338742743786,
"calibration/coverage@25%": 0.16337765274977642,
"calibration/coverage@30%": 0.32780288870889873,
"calibration/coverage@5%": 0.003938597405404571,
"calibration/distribution_entropy_10": 0.9300342218106076,
"calibration/distribution_entropy_100": 0.9450333391511456,
"calibration/ece": 0.16671802437009794,
"calibration/mean_confidence": 0.6319481031225281,
"calibration/unique_confidence_per_question": 0.994140625,
"calibration/unique_confidences": 509.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.004296875,
"completions/max_length": 1287.2,
"completions/max_terminated_length": 1287.2,
"completions/mean_length": 185.18955078125,
"completions/mean_terminated_length": 185.97795715332032,
"completions/min_length": 0.0,
"completions/min_terminated_length": 76.6,
"epoch": 0.16,
"grad_norm": 0.008226207457482815,
"learning_rate": 1e-06,
"loss": -0.0174,
"num_tokens": 165562528.0,
"reward": 0.9965122222900391,
"reward_std": 0.17300238013267516,
"rewards/accuracy_reward": 0.47392578125,
"rewards/batchaccgated_coverage_0": 0.1444621205329895,
"rewards/batchaccgated_coverage_1": 0.1444621205329895,
"rewards/batchaccgated_coverage_10": 0.19554198086261748,
"rewards/batchaccgated_coverage_15": 0.20201340317726135,
"rewards/batchaccgated_coverage_20": 0.21046102643013,
"rewards/batchaccgated_coverage_25": 0.21625565886497497,
"rewards/batchaccgated_coverage_5": 0.17266225516796113,
"rewards/brier_reward": 0.7324449181556701,
"rewards/confidence_uniqueness_reward": 0.9365719795227051,
"rewards/format_reward": 0.9955078125,
"rewards/frontier_entropy_batch_reward": -0.33692111968994143,
"signal/accuracy_reward/centered_abs_mean": 0.147222900390625,
"signal/accuracy_reward/group_std_mean": 0.18752501904964447,
"signal/accuracy_reward/group_zero_std_frac": 0.49375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.5907938838005066,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0736114501953125,
"signal/advantage_abs_mean": 0.7649896502494812,
"signal/advantage_pre_scale_abs_mean": 0.1314418524503708,
"signal/advantage_pre_scale_std": 0.19012850224971772,
"signal/advantage_std": 0.9836825728416443,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.06562154814600944,
"signal/batchaccgated_coverage_0/group_std_mean": 0.08731849193572998,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.05292134433984756,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.006562154460698366,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.06562154814600944,
"signal/batchaccgated_coverage_1/group_std_mean": 0.08731849193572998,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.05292134433984756,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.006562154460698366,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.07958894968032837,
"signal/batchaccgated_coverage_10/group_std_mean": 0.10361521989107132,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06469270884990692,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.007958895061165094,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.08306524455547333,
"signal/batchaccgated_coverage_15/group_std_mean": 0.1078745499253273,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06735358983278275,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.008306524716317654,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.0884243443608284,
"signal/batchaccgated_coverage_20/group_std_mean": 0.11363146007061005,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.07168330550193787,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.00884243445470929,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.09462088495492935,
"signal/batchaccgated_coverage_25/group_std_mean": 0.12070343643426895,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.07710714936256409,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.009462088346481323,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.07232067734003067,
"signal/batchaccgated_coverage_5/group_std_mean": 0.09503661692142487,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.05854475647211075,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.007232067827135325,
"signal/brier_reward/centered_abs_mean": 0.20488883256912233,
"signal/brier_reward/group_std_mean": 0.2545566469430923,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1671787530183792,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.020488884299993515,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.030331097170710565,
"signal/confidence_uniqueness_reward/group_std_mean": 0.049241021275520325,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.025027815625071524,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0030331098940223457,
"signal/format_reward/centered_abs_mean": 0.00863037109375,
"signal/format_reward/group_std_mean": 0.023670893907546998,
"signal/format_reward/group_zero_std_frac": 0.871875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.03370587155222893,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.004315185546875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.38159212470054626,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.4508224368095398,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.31222112774848937,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.03815921321511269,
"step": 50
},
{
"epoch": 0.16,
"eval_completions/clipped_ratio": 0.005859375,
"eval_completions/max_length": 519.0,
"eval_completions/max_terminated_length": 519.0,
"eval_completions/mean_length": 204.35829544067383,
"eval_completions/mean_terminated_length": 205.55280685424805,
"eval_completions/min_length": 27.25,
"eval_completions/min_terminated_length": 100.0,
"eval_loss": 0.0,
"eval_num_tokens": 165562528.0,
"eval_reward": 0.7513801157474518,
"eval_reward_std": 0.2504737749695778,
"eval_rewards/accuracy_reward": 0.369140625,
"eval_rewards/batchaccgated_coverage_0": 0.017853989731520414,
"eval_rewards/batchaccgated_coverage_1": 0.017853989731520414,
"eval_rewards/batchaccgated_coverage_10": 0.017853989731520414,
"eval_rewards/batchaccgated_coverage_15": 0.017344146966934204,
"eval_rewards/batchaccgated_coverage_20": 0.017023784574121237,
"eval_rewards/batchaccgated_coverage_25": 0.01641367725096643,
"eval_rewards/batchaccgated_coverage_5": 0.017853989731520414,
"eval_rewards/brier_reward": 0.7068519443273544,
"eval_rewards/confidence_uniqueness_reward": 0.8781108409166336,
"eval_rewards/format_reward": 0.990234375,
"eval_rewards/frontier_entropy_batch_reward": -0.990234375,
"eval_runtime": 47.2731,
"eval_samples_per_second": 10.577,
"eval_signal/accuracy_reward/centered_abs_mean": 0.4503173828125,
"eval_signal/accuracy_reward/group_std_mean": 0.48132171481847763,
"eval_signal/accuracy_reward/group_zero_std_frac": 0.0,
"eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9040320813655853,
"eval_signal/accuracy_reward/weight": 0.5,
"eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.22515869140625,
"eval_signal/advantage_abs_mean": 0.8881122171878815,
"eval_signal/advantage_pre_scale_abs_mean": 0.2226143553853035,
"eval_signal/advantage_pre_scale_std": 0.24828217178583145,
"eval_signal/advantage_std": 0.9877204895019531,
"eval_signal/batchaccgated_coverage_0/centered_abs_mean": 0.03536086808890104,
"eval_signal/batchaccgated_coverage_0/group_std_mean": 0.04805386811494827,
"eval_signal/batchaccgated_coverage_0/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.0141398583073169,
"eval_signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.003536087111569941,
"eval_signal/batchaccgated_coverage_1/centered_abs_mean": 0.03536086808890104,
"eval_signal/batchaccgated_coverage_1/group_std_mean": 0.04805386811494827,
"eval_signal/batchaccgated_coverage_1/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.0141398583073169,
"eval_signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.003536087111569941,
"eval_signal/batchaccgated_coverage_10/centered_abs_mean": 0.03536086808890104,
"eval_signal/batchaccgated_coverage_10/group_std_mean": 0.04805386811494827,
"eval_signal/batchaccgated_coverage_10/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.0141398583073169,
"eval_signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.003536087111569941,
"eval_signal/batchaccgated_coverage_15/centered_abs_mean": 0.03473409451544285,
"eval_signal/batchaccgated_coverage_15/group_std_mean": 0.04735161364078522,
"eval_signal/batchaccgated_coverage_15/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.013899537501856685,
"eval_signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0034734096261672676,
"eval_signal/batchaccgated_coverage_20/centered_abs_mean": 0.03402520064264536,
"eval_signal/batchaccgated_coverage_20/group_std_mean": 0.04649225063621998,
"eval_signal/batchaccgated_coverage_20/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.013623810140416026,
"eval_signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.003402520262170583,
"eval_signal/batchaccgated_coverage_25/centered_abs_mean": 0.03242296539247036,
"eval_signal/batchaccgated_coverage_25/group_std_mean": 0.04452566709369421,
"eval_signal/batchaccgated_coverage_25/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.013000429375097156,
"eval_signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.0032422966905869544,
"eval_signal/batchaccgated_coverage_5/centered_abs_mean": 0.03536086808890104,
"eval_signal/batchaccgated_coverage_5/group_std_mean": 0.04805386811494827,
"eval_signal/batchaccgated_coverage_5/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.0141398583073169,
"eval_signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003536087111569941,
"eval_signal/brier_reward/centered_abs_mean": 0.27812090516090393,
"eval_signal/brier_reward/group_std_mean": 0.3276386931538582,
"eval_signal/brier_reward/group_zero_std_frac": 0.0,
"eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.11172622442245483,
"eval_signal/brier_reward/weight": 0.10000000149011612,
"eval_signal/brier_reward/weighted_centered_abs_mean": 0.02781209209933877,
"eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.05934667959809303,
"eval_signal/confidence_uniqueness_reward/group_std_mean": 0.08842949755489826,
"eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.02371257171034813,
"eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.005934668006375432,
"eval_signal/format_reward/centered_abs_mean": 0.0186767578125,
"eval_signal/format_reward/group_std_mean": 0.04851654730737209,
"eval_signal/format_reward/group_zero_std_frac": 0.75,
"eval_signal/format_reward/scaled_weighted_centered_abs_mean": 0.03634009836241603,
"eval_signal/format_reward/weight": 0.5,
"eval_signal/format_reward/weighted_centered_abs_mean": 0.00933837890625,
"eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0186767578125,
"eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.04851654730737209,
"eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.75,
"eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.007268020184710622,
"eval_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"eval_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0018676758336368948,
"eval_steps_per_second": 0.085,
"step": 50
},
{
"epoch": 0.16,
"step": 50,
"train_probe_completions/clipped_ratio": 0.0,
"train_probe_completions/max_length": 575.75,
"train_probe_completions/max_terminated_length": 575.75,
"train_probe_completions/mean_length": 196.16736221313477,
"train_probe_completions/mean_terminated_length": 196.16736221313477,
"train_probe_completions/min_length": 92.25,
"train_probe_completions/min_terminated_length": 92.25,
"train_probe_loss": 0.0,
"train_probe_num_tokens": 165562528.0,
"train_probe_reward": 0.8221436440944672,
"train_probe_reward_std": 0.24294745922088623,
"train_probe_rewards/accuracy_reward": 0.48828125,
"train_probe_rewards/batchaccgated_coverage_0": 0.023933340329676867,
"train_probe_rewards/batchaccgated_coverage_1": 0.023933340329676867,
"train_probe_rewards/batchaccgated_coverage_10": 0.022934088949114084,
"train_probe_rewards/batchaccgated_coverage_15": 0.022063777316361666,
"train_probe_rewards/batchaccgated_coverage_20": 0.02106133010238409,
"train_probe_rewards/batchaccgated_coverage_25": 0.01987303141504526,
"train_probe_rewards/batchaccgated_coverage_5": 0.023933340329676867,
"train_probe_rewards/brier_reward": 0.7438799440860748,
"train_probe_rewards/confidence_uniqueness_reward": 0.87841796875,
"train_probe_rewards/format_reward": 1.0,
"train_probe_rewards/frontier_entropy_batch_reward": -1.0,
"train_probe_runtime": 25.474,
"train_probe_samples_per_second": 19.628,
"train_probe_signal/accuracy_reward/centered_abs_mean": 0.493896484375,
"train_probe_signal/accuracy_reward/group_std_mean": 0.5048737227916718,
"train_probe_signal/accuracy_reward/group_zero_std_frac": 0.0,
"train_probe_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.017445296049118,
"train_probe_signal/accuracy_reward/weight": 0.5,
"train_probe_signal/accuracy_reward/weighted_centered_abs_mean": 0.2469482421875,
"train_probe_signal/advantage_abs_mean": 0.9374572038650513,
"train_probe_signal/advantage_pre_scale_abs_mean": 0.22792625054717064,
"train_probe_signal/advantage_pre_scale_std": 0.24013962969183922,
"train_probe_signal/advantage_std": 0.9877113550901413,
"train_probe_signal/batchaccgated_coverage_0/centered_abs_mean": 0.05635852366685867,
"train_probe_signal/batchaccgated_coverage_0/group_std_mean": 0.07829220872372389,
"train_probe_signal/batchaccgated_coverage_0/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.023317129584029317,
"train_probe_signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.005635852401610464,
"train_probe_signal/batchaccgated_coverage_1/centered_abs_mean": 0.05635852366685867,
"train_probe_signal/batchaccgated_coverage_1/group_std_mean": 0.07829220872372389,
"train_probe_signal/batchaccgated_coverage_1/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.023317129584029317,
"train_probe_signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.005635852401610464,
"train_probe_signal/batchaccgated_coverage_10/centered_abs_mean": 0.05413336772471666,
"train_probe_signal/batchaccgated_coverage_10/group_std_mean": 0.07569092977792025,
"train_probe_signal/batchaccgated_coverage_10/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.022386874770745635,
"train_probe_signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.005413336853962392,
"train_probe_signal/batchaccgated_coverage_15/centered_abs_mean": 0.052296326495707035,
"train_probe_signal/batchaccgated_coverage_15/group_std_mean": 0.07344033941626549,
"train_probe_signal/batchaccgated_coverage_15/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.021633335389196873,
"train_probe_signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.005229632661212236,
"train_probe_signal/batchaccgated_coverage_20/centered_abs_mean": 0.049386327154934406,
"train_probe_signal/batchaccgated_coverage_20/group_std_mean": 0.0697292611002922,
"train_probe_signal/batchaccgated_coverage_20/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.020430967677384615,
"train_probe_signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.00493863265728578,
"train_probe_signal/batchaccgated_coverage_25/centered_abs_mean": 0.04411026602610946,
"train_probe_signal/batchaccgated_coverage_25/group_std_mean": 0.0621018810197711,
"train_probe_signal/batchaccgated_coverage_25/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.018241578713059425,
"train_probe_signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.0044110267772339284,
"train_probe_signal/batchaccgated_coverage_5/centered_abs_mean": 0.05635852366685867,
"train_probe_signal/batchaccgated_coverage_5/group_std_mean": 0.07829220872372389,
"train_probe_signal/batchaccgated_coverage_5/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.023317129584029317,
"train_probe_signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.005635852401610464,
"train_probe_signal/brier_reward/centered_abs_mean": 0.2579937316477299,
"train_probe_signal/brier_reward/group_std_mean": 0.31405670940876007,
"train_probe_signal/brier_reward/group_zero_std_frac": 0.0,
"train_probe_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.10614863969385624,
"train_probe_signal/brier_reward/weight": 0.10000000149011612,
"train_probe_signal/brier_reward/weighted_centered_abs_mean": 0.0257993726991117,
"train_probe_signal/confidence_uniqueness_reward/centered_abs_mean": 0.0537261962890625,
"train_probe_signal/confidence_uniqueness_reward/group_std_mean": 0.06682171579450369,
"train_probe_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"train_probe_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.022080628667026758,
"train_probe_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"train_probe_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.005372619605623186,
"train_probe_signal/format_reward/centered_abs_mean": 0.0,
"train_probe_signal/format_reward/group_std_mean": 0.0,
"train_probe_signal/format_reward/group_zero_std_frac": 1.0,
"train_probe_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/format_reward/weight": 0.5,
"train_probe_signal/format_reward/weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"train_probe_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"train_probe_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"train_probe_steps_per_second": 0.157
},
{
"calibration/aurc": 0.34774440171216764,
"calibration/batch_distribution_entropy": 0.9016706264351999,
"calibration/batch_entropy_100bins": 0.9273505918979653,
"calibration/batch_entropy_10bins": 0.9016706264351999,
"calibration/batch_entropy_50bins": 0.9283601141567841,
"calibration/batch_uniqueness": 0.9333117653896842,
"calibration/buffer_distribution_entropy": 0.9286699980517102,
"calibration/buffer_entropy_100bins": 0.9072736142956191,
"calibration/buffer_entropy_10bins": 0.9286699980517102,
"calibration/buffer_entropy_50bins": 0.925909761571179,
"calibration/confidence_entropy": 0.41494209800708803,
"calibration/coverage@0%": 0.003520211594911937,
"calibration/coverage@1%": 0.003520211594911937,
"calibration/coverage@10%": 0.012907442514677103,
"calibration/coverage@15%": 0.02621468321917808,
"calibration/coverage@20%": 0.046164077788649706,
"calibration/coverage@25%": 0.2586977128180039,
"calibration/coverage@30%": 0.46251352595832856,
"calibration/coverage@5%": 0.003520211594911937,
"calibration/distribution_entropy_10": 0.9016706264351999,
"calibration/distribution_entropy_100": 0.9273505918979653,
"calibration/ece": 0.1747755092990943,
"calibration/mean_confidence": 0.6607646708340997,
"calibration/unique_confidence_per_question": 0.998828125,
"calibration/unique_confidences": 511.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0017578125,
"completions/max_length": 1069.6,
"completions/max_terminated_length": 1069.6,
"completions/mean_length": 200.9591796875,
"completions/mean_terminated_length": 201.31504516601564,
"completions/min_length": 17.4,
"completions/min_terminated_length": 93.0,
"epoch": 0.176,
"grad_norm": 0.009802790358662605,
"learning_rate": 1e-06,
"loss": -0.0078,
"num_tokens": 182857470.0,
"reward": 1.0193021416664123,
"reward_std": 0.15085843205451965,
"rewards/accuracy_reward": 0.48330078125,
"rewards/batchaccgated_coverage_0": 0.16325082033872604,
"rewards/batchaccgated_coverage_1": 0.16325082033872604,
"rewards/batchaccgated_coverage_10": 0.2175288736820221,
"rewards/batchaccgated_coverage_15": 0.22109545767307281,
"rewards/batchaccgated_coverage_20": 0.23151999711990356,
"rewards/batchaccgated_coverage_25": 0.23789876997470855,
"rewards/batchaccgated_coverage_5": 0.18858735859394074,
"rewards/brier_reward": 0.742801308631897,
"rewards/confidence_uniqueness_reward": 0.935167932510376,
"rewards/format_reward": 0.99814453125,
"rewards/frontier_entropy_batch_reward": -0.31530646681785585,
"signal/accuracy_reward/centered_abs_mean": 0.123651123046875,
"signal/accuracy_reward/group_std_mean": 0.16307915449142457,
"signal/accuracy_reward/group_zero_std_frac": 0.5375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.6281829237937927,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0618255615234375,
"signal/advantage_abs_mean": 0.7436140298843383,
"signal/advantage_pre_scale_abs_mean": 0.11255786716938018,
"signal/advantage_pre_scale_std": 0.16903536915779113,
"signal/advantage_std": 0.9834771275520324,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.060704483091831206,
"signal/batchaccgated_coverage_0/group_std_mean": 0.07957982569932938,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06253724396228791,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.006070448458194733,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.060704483091831206,
"signal/batchaccgated_coverage_1/group_std_mean": 0.07957982569932938,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06253724396228791,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.006070448458194733,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.07631024718284607,
"signal/batchaccgated_coverage_10/group_std_mean": 0.10019671022891999,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07795228362083435,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0076310249976813795,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.07726057469844819,
"signal/batchaccgated_coverage_15/group_std_mean": 0.10093590915203095,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.07889396101236343,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.00772605761885643,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.08439825475215912,
"signal/batchaccgated_coverage_20/group_std_mean": 0.11011920422315598,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.08613496869802476,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.008439825475215912,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.09127712398767471,
"signal/batchaccgated_coverage_25/group_std_mean": 0.11843348890542985,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.0931567445397377,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.00912771262228489,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.06640159115195274,
"signal/batchaccgated_coverage_5/group_std_mean": 0.08670850247144699,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06783257871866226,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.006640159152448177,
"signal/brier_reward/centered_abs_mean": 0.17577806413173674,
"signal/brier_reward/group_std_mean": 0.2228584885597229,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.18079962134361266,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.01757780686020851,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.030004977807402612,
"signal/confidence_uniqueness_reward/group_std_mean": 0.04296218454837799,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.03080202415585518,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0030004978645592926,
"signal/format_reward/centered_abs_mean": 0.003582763671875,
"signal/format_reward/group_std_mean": 0.010159807885065675,
"signal/format_reward/group_zero_std_frac": 0.94375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.01898806421086192,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0017913818359375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.34557713866233825,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.42060938477516174,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.35595895648002623,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.034557714313268664,
"step": 55
},
{
"calibration/aurc": 0.26461575631539896,
"calibration/batch_distribution_entropy": 0.8762495962411314,
"calibration/batch_entropy_100bins": 0.9014260710837316,
"calibration/batch_entropy_10bins": 0.8762495962411314,
"calibration/batch_entropy_50bins": 0.899305163833289,
"calibration/batch_uniqueness": 0.9101348876953125,
"calibration/buffer_distribution_entropy": 0.9280459863748348,
"calibration/buffer_entropy_100bins": 0.915551232854947,
"calibration/buffer_entropy_10bins": 0.9280459863748348,
"calibration/buffer_entropy_50bins": 0.9303390097557239,
"calibration/confidence_entropy": 0.37436898873292634,
"calibration/coverage@0%": 0.0046875,
"calibration/coverage@1%": 0.0046875,
"calibration/coverage@10%": 0.13671875,
"calibration/coverage@15%": 0.241015625,
"calibration/coverage@20%": 0.394140625,
"calibration/coverage@25%": 0.4828125,
"calibration/coverage@30%": 0.629296875,
"calibration/coverage@5%": 0.06015625,
"calibration/distribution_entropy_10": 0.8762495962411314,
"calibration/distribution_entropy_100": 0.9014260710837316,
"calibration/ece": 0.14908437310937134,
"calibration/mean_confidence": 0.6426375267306124,
"calibration/unique_confidence_per_question": 1.0,
"calibration/unique_confidences": 512.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 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": 799.8,
"completions/max_terminated_length": 799.8,
"completions/mean_length": 199.53759765625,
"completions/mean_terminated_length": 199.67336730957032,
"completions/min_length": 19.0,
"completions/min_terminated_length": 95.2,
"epoch": 0.192,
"grad_norm": 0.008111663162708282,
"learning_rate": 1e-06,
"loss": 0.0023,
"num_tokens": 199715551.0,
"reward": 1.0609397411346435,
"reward_std": 0.145148167014122,
"rewards/accuracy_reward": 0.51474609375,
"rewards/batchaccgated_coverage_0": 0.18850284814834595,
"rewards/batchaccgated_coverage_1": 0.18850284814834595,
"rewards/batchaccgated_coverage_10": 0.2590787947177887,
"rewards/batchaccgated_coverage_15": 0.26692615151405336,
"rewards/batchaccgated_coverage_20": 0.27537758350372316,
"rewards/batchaccgated_coverage_25": 0.28158064782619474,
"rewards/batchaccgated_coverage_5": 0.22956853210926056,
"rewards/brier_reward": 0.7692630410194397,
"rewards/confidence_uniqueness_reward": 0.9096927404403686,
"rewards/format_reward": 0.99921875,
"rewards/frontier_entropy_batch_reward": -0.32892040014266966,
"signal/accuracy_reward/centered_abs_mean": 0.117535400390625,
"signal/accuracy_reward/group_std_mean": 0.15951007902622222,
"signal/accuracy_reward/group_zero_std_frac": 0.53125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.797590184211731,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0587677001953125,
"signal/advantage_abs_mean": 0.7195308804512024,
"signal/advantage_pre_scale_abs_mean": 0.10562613159418106,
"signal/advantage_pre_scale_std": 0.16929024159908296,
"signal/advantage_std": 0.9831543684005737,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.05814865082502365,
"signal/batchaccgated_coverage_0/group_std_mean": 0.07534236162900924,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07802018225193023,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.005814864858984947,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.05814865082502365,
"signal/batchaccgated_coverage_1/group_std_mean": 0.07534236162900924,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07802018225193023,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.005814864858984947,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.0781184583902359,
"signal/batchaccgated_coverage_10/group_std_mean": 0.10500485002994538,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.10490109771490097,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.007811845932155848,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.08099955320358276,
"signal/batchaccgated_coverage_15/group_std_mean": 0.10923604220151902,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.10896416157484054,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.008099955786019564,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.08582237809896469,
"signal/batchaccgated_coverage_20/group_std_mean": 0.11549726575613022,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.11529813855886459,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.00858223820105195,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.09185716062784195,
"signal/batchaccgated_coverage_25/group_std_mean": 0.12292417138814926,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.12325630486011505,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.00918571585789323,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.06898609027266503,
"signal/batchaccgated_coverage_5/group_std_mean": 0.09091597199440002,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.290625,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.09272701442241668,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.006898609269410372,
"signal/brier_reward/centered_abs_mean": 0.1545909136533737,
"signal/brier_reward/group_std_mean": 0.20077145993709564,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.21058568358421326,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.01545909084379673,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.04500208981335163,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06004456654191017,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.060928435623645784,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004500208981335163,
"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": 0.009725516941398383,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000750732421875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.3003262996673584,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.38225622177124025,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.015625,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.40684792399406433,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.030032632127404212,
"step": 60
},
{
"calibration/aurc": 0.2371006506091347,
"calibration/batch_distribution_entropy": 0.7332016981392208,
"calibration/batch_entropy_100bins": 0.7877366787187977,
"calibration/batch_entropy_10bins": 0.7332016981392208,
"calibration/batch_entropy_50bins": 0.7702146412414355,
"calibration/batch_uniqueness": 0.7965023524984324,
"calibration/buffer_distribution_entropy": 0.9215342601032634,
"calibration/buffer_entropy_100bins": 0.9165937111515241,
"calibration/buffer_entropy_10bins": 0.9215342601032634,
"calibration/buffer_entropy_50bins": 0.9278052650065092,
"calibration/confidence_entropy": 0.2985602625577954,
"calibration/coverage@0%": 0.009767918297455968,
"calibration/coverage@1%": 0.009767918297455968,
"calibration/coverage@10%": 0.18908008194716244,
"calibration/coverage@15%": 0.38855644569471626,
"calibration/coverage@20%": 0.5734183891878669,
"calibration/coverage@25%": 0.625,
"calibration/coverage@30%": 0.66953125,
"calibration/coverage@5%": 0.009767918297455968,
"calibration/distribution_entropy_10": 0.7332016981392208,
"calibration/distribution_entropy_100": 0.7877366787187977,
"calibration/ece": 0.14922211267473307,
"calibration/mean_confidence": 0.7332826692663694,
"calibration/unique_confidence_per_question": 0.999609375,
"calibration/unique_confidences": 511.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00048828125,
"completions/max_length": 790.4,
"completions/max_terminated_length": 790.4,
"completions/mean_length": 197.0931640625,
"completions/mean_terminated_length": 197.1896240234375,
"completions/min_length": 17.6,
"completions/min_terminated_length": 88.4,
"epoch": 0.208,
"grad_norm": 0.007880998775362968,
"learning_rate": 1e-06,
"loss": -0.0015,
"num_tokens": 216766009.0,
"reward": 1.1107650995254517,
"reward_std": 0.1430345892906189,
"rewards/accuracy_reward": 0.55400390625,
"rewards/batchaccgated_coverage_0": 0.25276653468608856,
"rewards/batchaccgated_coverage_1": 0.25276653468608856,
"rewards/batchaccgated_coverage_10": 0.3330254673957825,
"rewards/batchaccgated_coverage_15": 0.34098418354988097,
"rewards/batchaccgated_coverage_20": 0.3478421032428741,
"rewards/batchaccgated_coverage_25": 0.35393781661987306,
"rewards/batchaccgated_coverage_5": 0.29933540225028993,
"rewards/brier_reward": 0.7631839871406555,
"rewards/confidence_uniqueness_reward": 0.812430202960968,
"rewards/format_reward": 0.99931640625,
"rewards/frontier_entropy_batch_reward": -0.41522324085235596,
"signal/accuracy_reward/centered_abs_mean": 0.115802001953125,
"signal/accuracy_reward/group_std_mean": 0.15337497293949126,
"signal/accuracy_reward/group_zero_std_frac": 0.55625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.1290308594703675,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0579010009765625,
"signal/advantage_abs_mean": 0.7050060987472534,
"signal/advantage_pre_scale_abs_mean": 0.10349384993314743,
"signal/advantage_pre_scale_std": 0.17194433212280275,
"signal/advantage_std": 0.9825743556022644,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.05089330375194549,
"signal/batchaccgated_coverage_0/group_std_mean": 0.07049596160650254,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.2625,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.09692869037389755,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.005089330393821001,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.05089330375194549,
"signal/batchaccgated_coverage_1/group_std_mean": 0.07049596160650254,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.2625,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.09692869037389755,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.005089330393821001,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.07291491180658341,
"signal/batchaccgated_coverage_10/group_std_mean": 0.10234667360782623,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.2625,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.13932026028633118,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.007291491702198982,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.07711826711893081,
"signal/batchaccgated_coverage_15/group_std_mean": 0.1078446552157402,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.2625,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.14763022363185882,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.007711827103048563,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.08296172618865967,
"signal/batchaccgated_coverage_20/group_std_mean": 0.11500341147184372,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.2625,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.159185528755188,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.00829617278650403,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.09045031219720841,
"signal/batchaccgated_coverage_25/group_std_mean": 0.12373146712779999,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.2625,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.17463753819465638,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.009045031480491161,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.06173686161637306,
"signal/batchaccgated_coverage_5/group_std_mean": 0.0861477792263031,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.2625,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.11705326288938522,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.006173686310648918,
"signal/brier_reward/centered_abs_mean": 0.1565554767847061,
"signal/brier_reward/group_std_mean": 0.20119291841983794,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.3064535647630692,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.015655547380447388,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.10816548019647598,
"signal/confidence_uniqueness_reward/group_std_mean": 0.13739174008369445,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.21456506848335266,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.010816548205912112,
"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": 0.012464709766209125,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0006622314453125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.30826908349990845,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.392506992816925,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0375,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5962862372398376,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.030826908349990845,
"step": 65
},
{
"calibration/aurc": 0.24884904269284602,
"calibration/batch_distribution_entropy": 0.8306280055696128,
"calibration/batch_entropy_100bins": 0.8252312780203589,
"calibration/batch_entropy_10bins": 0.8306280055696128,
"calibration/batch_entropy_50bins": 0.825268780426929,
"calibration/batch_uniqueness": 0.8327884132943051,
"calibration/buffer_distribution_entropy": 0.9154187936655755,
"calibration/buffer_entropy_100bins": 0.9148061173775671,
"calibration/buffer_entropy_10bins": 0.9154187936655755,
"calibration/buffer_entropy_50bins": 0.9236344454847514,
"calibration/confidence_entropy": 0.3199919289629761,
"calibration/coverage@0%": 0.034765625,
"calibration/coverage@1%": 0.034765625,
"calibration/coverage@10%": 0.31640625,
"calibration/coverage@15%": 0.41328125,
"calibration/coverage@20%": 0.47734375,
"calibration/coverage@25%": 0.540625,
"calibration/coverage@30%": 0.609375,
"calibration/coverage@5%": 0.22109375,
"calibration/distribution_entropy_10": 0.8306280055696128,
"calibration/distribution_entropy_100": 0.8252312780203589,
"calibration/ece": 0.1684312866396553,
"calibration/mean_confidence": 0.6587055922340608,
"calibration/unique_confidence_per_question": 0.998828125,
"calibration/unique_confidences": 511.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00087890625,
"completions/max_length": 925.0,
"completions/max_terminated_length": 925.0,
"completions/mean_length": 194.01826171875,
"completions/mean_terminated_length": 194.18970336914063,
"completions/min_length": 37.2,
"completions/min_terminated_length": 96.6,
"epoch": 0.224,
"grad_norm": 0.008232714608311653,
"learning_rate": 1e-06,
"loss": -0.0033,
"num_tokens": 233905940.0,
"reward": 1.074370813369751,
"reward_std": 0.12654837369918823,
"rewards/accuracy_reward": 0.496875,
"rewards/batchaccgated_coverage_0": 0.2565078675746918,
"rewards/batchaccgated_coverage_1": 0.2565078675746918,
"rewards/batchaccgated_coverage_10": 0.3127350687980652,
"rewards/batchaccgated_coverage_15": 0.31579423546791074,
"rewards/batchaccgated_coverage_20": 0.32091079354286195,
"rewards/batchaccgated_coverage_25": 0.32387442588806153,
"rewards/batchaccgated_coverage_5": 0.2870495796203613,
"rewards/brier_reward": 0.7683147192001343,
"rewards/confidence_uniqueness_reward": 0.8057521343231201,
"rewards/format_reward": 0.9990234375,
"rewards/frontier_entropy_batch_reward": -0.3832306981086731,
"signal/accuracy_reward/centered_abs_mean": 0.109375,
"signal/accuracy_reward/group_std_mean": 0.13956573605537415,
"signal/accuracy_reward/group_zero_std_frac": 0.61875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.1803191661834718,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0546875,
"signal/advantage_abs_mean": 0.6982428669929505,
"signal/advantage_pre_scale_abs_mean": 0.09274145662784576,
"signal/advantage_pre_scale_std": 0.1560712069272995,
"signal/advantage_std": 0.9823532581329346,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.04054286181926727,
"signal/batchaccgated_coverage_0/group_std_mean": 0.05448315069079399,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.325,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08895989432930947,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.004054286330938339,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.04054286181926727,
"signal/batchaccgated_coverage_1/group_std_mean": 0.05448315069079399,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.325,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08895989432930947,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.004054286330938339,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.054483331739902496,
"signal/batchaccgated_coverage_10/group_std_mean": 0.07707191407680511,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.325,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.12059174180030822,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.005448333453387022,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.05553565993905067,
"signal/batchaccgated_coverage_15/group_std_mean": 0.07815761268138885,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.325,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.12240631580352783,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.005553566012531519,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.05919899269938469,
"signal/batchaccgated_coverage_20/group_std_mean": 0.08315939754247666,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.325,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.13104261457920074,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0059198994189500805,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.06282573640346527,
"signal/batchaccgated_coverage_25/group_std_mean": 0.08744254857301711,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.325,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.1378329277038574,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.006282573845237493,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.04681948646903038,
"signal/batchaccgated_coverage_5/group_std_mean": 0.06427032276988029,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.325,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.10279254466295243,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.004681948851794004,
"signal/brier_reward/centered_abs_mean": 0.14869073331356047,
"signal/brier_reward/group_std_mean": 0.18988703787326813,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.3224425345659256,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.014869073405861855,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.1067460373044014,
"signal/confidence_uniqueness_reward/group_std_mean": 0.13690947443246843,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.23263682574033737,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.010674603935331105,
"signal/format_reward/centered_abs_mean": 0.00185546875,
"signal/format_reward/group_std_mean": 0.004792441846802831,
"signal/format_reward/group_zero_std_frac": 0.975,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.02102384939789772,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000927734375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28738937973976136,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.37481691837310793,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.03125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.6271644055843353,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02873893976211548,
"step": 70
},
{
"calibration/aurc": 0.31671302994897665,
"calibration/batch_distribution_entropy": 0.8518835771388147,
"calibration/batch_entropy_100bins": 0.87090569729219,
"calibration/batch_entropy_10bins": 0.8518835771388147,
"calibration/batch_entropy_50bins": 0.8707932188655818,
"calibration/batch_uniqueness": 0.8936263530571861,
"calibration/buffer_distribution_entropy": 0.9135504589629762,
"calibration/buffer_entropy_100bins": 0.9155478445301608,
"calibration/buffer_entropy_10bins": 0.9135504589629762,
"calibration/buffer_entropy_50bins": 0.922841730704351,
"calibration/confidence_entropy": 0.36680822519450523,
"calibration/coverage@0%": 0.028517153864970647,
"calibration/coverage@1%": 0.028517153864970647,
"calibration/coverage@10%": 0.15664215386497066,
"calibration/coverage@15%": 0.19804840386497063,
"calibration/coverage@20%": 0.26017536081213305,
"calibration/coverage@25%": 0.31369404354207436,
"calibration/coverage@30%": 0.5125236974070451,
"calibration/coverage@5%": 0.10586090386497064,
"calibration/distribution_entropy_10": 0.8518835771388147,
"calibration/distribution_entropy_100": 0.87090569729219,
"calibration/ece": 0.18168518895018806,
"calibration/mean_confidence": 0.6667802928721105,
"calibration/unique_confidence_per_question": 0.994140625,
"calibration/unique_confidences": 509.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.000390625,
"completions/max_length": 557.4,
"completions/max_terminated_length": 557.4,
"completions/mean_length": 196.89140625,
"completions/mean_terminated_length": 196.96788940429687,
"completions/min_length": 29.2,
"completions/min_terminated_length": 91.6,
"epoch": 0.24,
"grad_norm": 0.008408453315496445,
"learning_rate": 1e-06,
"loss": 0.0002,
"num_tokens": 251173788.0,
"reward": 1.1068171739578248,
"reward_std": 0.13669639825820923,
"rewards/accuracy_reward": 0.54755859375,
"rewards/batchaccgated_coverage_0": 0.2486361861228943,
"rewards/batchaccgated_coverage_1": 0.2486361861228943,
"rewards/batchaccgated_coverage_10": 0.3067008137702942,
"rewards/batchaccgated_coverage_15": 0.31257139444351195,
"rewards/batchaccgated_coverage_20": 0.31476805806159974,
"rewards/batchaccgated_coverage_25": 0.31924829185009,
"rewards/batchaccgated_coverage_5": 0.28447783887386324,
"rewards/brier_reward": 0.7745516419410705,
"rewards/confidence_uniqueness_reward": 0.8909048318862915,
"rewards/format_reward": 0.9994140625,
"rewards/frontier_entropy_batch_reward": -0.3671866893768311,
"signal/accuracy_reward/centered_abs_mean": 0.121099853515625,
"signal/accuracy_reward/group_std_mean": 0.16128340661525725,
"signal/accuracy_reward/group_zero_std_frac": 0.53125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.032259726524353,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0605499267578125,
"signal/advantage_abs_mean": 0.7132783770561218,
"signal/advantage_pre_scale_abs_mean": 0.09960681200027466,
"signal/advantage_pre_scale_std": 0.16131314635276794,
"signal/advantage_std": 0.9827737808227539,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.050003471225500105,
"signal/batchaccgated_coverage_0/group_std_mean": 0.06580036208033561,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.221875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08660197407007217,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.005000347271561623,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.050003471225500105,
"signal/batchaccgated_coverage_1/group_std_mean": 0.06580036208033561,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.221875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08660197407007217,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.005000347271561623,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.06549834609031677,
"signal/batchaccgated_coverage_10/group_std_mean": 0.08970722854137421,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.221875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.11413258910179139,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.006549834553152323,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.06888469606637955,
"signal/batchaccgated_coverage_15/group_std_mean": 0.09407087117433548,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.221875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.12026008516550064,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.006888470053672791,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.06929319500923156,
"signal/batchaccgated_coverage_20/group_std_mean": 0.09455493241548538,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.221875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.12112376093864441,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.006929320003837347,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.0728753685951233,
"signal/batchaccgated_coverage_25/group_std_mean": 0.09922638237476349,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.221875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.12801474779844285,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.007287536840885877,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.05786111205816269,
"signal/batchaccgated_coverage_5/group_std_mean": 0.07823317646980285,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.221875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.10073961913585663,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.005786111392080784,
"signal/brier_reward/centered_abs_mean": 0.13829172551631927,
"signal/brier_reward/group_std_mean": 0.1781446099281311,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.23790936172008514,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.013829173147678375,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.05415357574820519,
"signal/confidence_uniqueness_reward/group_std_mean": 0.07144427746534347,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0959607869386673,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.005415357742458582,
"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": 0.009575171954929828,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000567626953125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2852127969264984,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3624271035194397,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.025,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.49137144684791567,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028521280735731125,
"step": 75
},
{
"calibration/aurc": 0.2336217676603131,
"calibration/batch_distribution_entropy": 0.741825986050106,
"calibration/batch_entropy_100bins": 0.78850994788832,
"calibration/batch_entropy_10bins": 0.741825986050106,
"calibration/batch_entropy_50bins": 0.7783956554809109,
"calibration/batch_uniqueness": 0.8189855134312074,
"calibration/buffer_distribution_entropy": 0.9084523816955224,
"calibration/buffer_entropy_100bins": 0.914060596409143,
"calibration/buffer_entropy_10bins": 0.9084523816955224,
"calibration/buffer_entropy_50bins": 0.9196756121182424,
"calibration/confidence_entropy": 0.30042311617148876,
"calibration/coverage@0%": 0.06328125,
"calibration/coverage@1%": 0.0703125,
"calibration/coverage@10%": 0.38125,
"calibration/coverage@15%": 0.42265625,
"calibration/coverage@20%": 0.46640625,
"calibration/coverage@25%": 0.509375,
"calibration/coverage@30%": 0.690234375,
"calibration/coverage@5%": 0.23125,
"calibration/distribution_entropy_10": 0.741825986050106,
"calibration/distribution_entropy_100": 0.78850994788832,
"calibration/ece": 0.16463053879078085,
"calibration/mean_confidence": 0.7439195033360081,
"calibration/unique_confidence_per_question": 0.962890625,
"calibration/unique_confidences": 493.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00029296875,
"completions/max_length": 703.8,
"completions/max_terminated_length": 703.8,
"completions/mean_length": 198.9806640625,
"completions/mean_terminated_length": 199.04030151367186,
"completions/min_length": 42.2,
"completions/min_terminated_length": 102.2,
"epoch": 0.256,
"grad_norm": 0.009298115968704224,
"learning_rate": 1e-06,
"loss": -0.0036,
"num_tokens": 268266166.0,
"reward": 1.1166703939437865,
"reward_std": 0.123066146671772,
"rewards/accuracy_reward": 0.54443359375,
"rewards/batchaccgated_coverage_0": 0.29064599275588987,
"rewards/batchaccgated_coverage_1": 0.29064599275588987,
"rewards/batchaccgated_coverage_10": 0.3461409032344818,
"rewards/batchaccgated_coverage_15": 0.34866883754730227,
"rewards/batchaccgated_coverage_20": 0.3591776490211487,
"rewards/batchaccgated_coverage_25": 0.367471444606781,
"rewards/batchaccgated_coverage_5": 0.32220990061759947,
"rewards/brier_reward": 0.7620518684387207,
"rewards/confidence_uniqueness_reward": 0.8020129203796387,
"rewards/format_reward": 0.999609375,
"rewards/frontier_entropy_batch_reward": -0.4425364375114441,
"signal/accuracy_reward/centered_abs_mean": 0.111663818359375,
"signal/accuracy_reward/group_std_mean": 0.14512262642383575,
"signal/accuracy_reward/group_zero_std_frac": 0.59375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 3.88775429725647,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0558319091796875,
"signal/advantage_abs_mean": 0.7022534728050231,
"signal/advantage_pre_scale_abs_mean": 0.090090049803257,
"signal/advantage_pre_scale_std": 0.15792504251003264,
"signal/advantage_std": 0.9794187664985656,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.03587271124124527,
"signal/batchaccgated_coverage_0/group_std_mean": 0.04879908785223961,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.26875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.21971785724163057,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0035872711334377526,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.03587271124124527,
"signal/batchaccgated_coverage_1/group_std_mean": 0.04879908785223961,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.26875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.21971785724163057,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0035872711334377526,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.05010294280946255,
"signal/batchaccgated_coverage_10/group_std_mean": 0.07199981287121773,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.26875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.3178232371807098,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.00501029440201819,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.05111348256468773,
"signal/batchaccgated_coverage_15/group_std_mean": 0.07331958711147309,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.26875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.3192783683538437,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.005111348070204258,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.05792687013745308,
"signal/batchaccgated_coverage_20/group_std_mean": 0.08246087431907653,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.26875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.3555942505598068,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.005792687181383371,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.06741169169545173,
"signal/batchaccgated_coverage_25/group_std_mean": 0.0939914345741272,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.26875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.4059398204088211,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.006741169281303883,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.04158569164574146,
"signal/batchaccgated_coverage_5/group_std_mean": 0.05838819891214371,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.26875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.25778297930955885,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.004158569499850273,
"signal/brier_reward/centered_abs_mean": 0.12922424972057342,
"signal/brier_reward/group_std_mean": 0.16605657041072847,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.8447867333889008,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.01292242519557476,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.09654060602188111,
"signal/confidence_uniqueness_reward/group_std_mean": 0.12155406177043915,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.003125,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.5835326045751572,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.009654060751199723,
"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": 0.030370459705591202,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.00037841796875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.250981280207634,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.32713971138000486,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0875,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.5988227605819703,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.025098129734396936,
"step": 80
},
{
"calibration/aurc": 0.3441383301461232,
"calibration/batch_distribution_entropy": 0.7837626024033365,
"calibration/batch_entropy_100bins": 0.8264899922042364,
"calibration/batch_entropy_10bins": 0.7837626024033365,
"calibration/batch_entropy_50bins": 0.8180669565229046,
"calibration/batch_uniqueness": 0.8604400634765625,
"calibration/buffer_distribution_entropy": 0.9022734721313996,
"calibration/buffer_entropy_100bins": 0.9113500834196101,
"calibration/buffer_entropy_10bins": 0.9022734721313996,
"calibration/buffer_entropy_50bins": 0.9153674786488064,
"calibration/confidence_entropy": 0.34929423998079,
"calibration/coverage@0%": 0.021875,
"calibration/coverage@1%": 0.021875,
"calibration/coverage@10%": 0.194921875,
"calibration/coverage@15%": 0.260546875,
"calibration/coverage@20%": 0.3109375,
"calibration/coverage@25%": 0.36015625,
"calibration/coverage@30%": 0.4375,
"calibration/coverage@5%": 0.103125,
"calibration/distribution_entropy_10": 0.7837626024033365,
"calibration/distribution_entropy_100": 0.8264899922042364,
"calibration/ece": 0.25224774449370735,
"calibration/mean_confidence": 0.720758860304036,
"calibration/unique_confidence_per_question": 0.96640625,
"calibration/unique_confidences": 494.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 586.0,
"completions/max_terminated_length": 586.0,
"completions/mean_length": 212.2359375,
"completions/mean_terminated_length": 212.25682678222657,
"completions/min_length": 80.4,
"completions/min_terminated_length": 102.0,
"epoch": 0.272,
"grad_norm": 0.008592949248850346,
"learning_rate": 1e-06,
"loss": 0.002,
"num_tokens": 285405158.0,
"reward": 1.0876873970031737,
"reward_std": 0.11875751316547394,
"rewards/accuracy_reward": 0.51640625,
"rewards/batchaccgated_coverage_0": 0.25274882912635804,
"rewards/batchaccgated_coverage_1": 0.25274882912635804,
"rewards/batchaccgated_coverage_10": 0.3109740734100342,
"rewards/batchaccgated_coverage_15": 0.3153887152671814,
"rewards/batchaccgated_coverage_20": 0.3270998537540436,
"rewards/batchaccgated_coverage_25": 0.33288330435752866,
"rewards/batchaccgated_coverage_5": 0.2790035307407379,
"rewards/brier_reward": 0.761996066570282,
"rewards/confidence_uniqueness_reward": 0.865186071395874,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.40220981240272524,
"signal/accuracy_reward/centered_abs_mean": 0.10323486328125,
"signal/accuracy_reward/group_std_mean": 0.1362008735537529,
"signal/accuracy_reward/group_zero_std_frac": 0.621875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 3.4977614879608154,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.051617431640625,
"signal/advantage_abs_mean": 0.7008754730224609,
"signal/advantage_pre_scale_abs_mean": 0.08583134859800338,
"signal/advantage_pre_scale_std": 0.15158269703388214,
"signal/advantage_std": 0.9794724345207214,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.02992990016937256,
"signal/batchaccgated_coverage_0/group_std_mean": 0.04212051779031754,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.334375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.17607344835996627,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0029929901007562875,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.02992990016937256,
"signal/batchaccgated_coverage_1/group_std_mean": 0.04212051779031754,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.334375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.17607344835996627,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0029929901007562875,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.046000920981168744,
"signal/batchaccgated_coverage_10/group_std_mean": 0.06815331056714058,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.334375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.25800175219774246,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004600092303007841,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.047893743962049484,
"signal/batchaccgated_coverage_15/group_std_mean": 0.0706049457192421,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.334375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.28227021396160124,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004789374396204948,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.0567482627928257,
"signal/batchaccgated_coverage_20/group_std_mean": 0.08234160169959068,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.334375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.3586976259946823,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.005674826260656118,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.0633957862854004,
"signal/batchaccgated_coverage_25/group_std_mean": 0.09037741124629975,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.334375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.4275981977581978,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.00633957888931036,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03442709147930145,
"signal/batchaccgated_coverage_5/group_std_mean": 0.05014726296067238,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.334375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.20199376344680786,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0034427092410624026,
"signal/brier_reward/centered_abs_mean": 0.12312300801277161,
"signal/brier_reward/group_std_mean": 0.16031525731086732,
"signal/brier_reward/group_zero_std_frac": 0.003125,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.8383482873439789,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.012312300875782967,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.056654618680477144,
"signal/confidence_uniqueness_reward/group_std_mean": 0.07380898296833038,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.003125,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.4246091768145561,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.005665462091565132,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.01910267248749733,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2511567234992981,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3292468786239624,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.084375,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.5229641318321228,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02511567361652851,
"step": 85
},
{
"calibration/aurc": 0.26413717707979556,
"calibration/batch_distribution_entropy": 0.8275800902466568,
"calibration/batch_entropy_100bins": 0.8529547150479395,
"calibration/batch_entropy_10bins": 0.8275800902466568,
"calibration/batch_entropy_50bins": 0.8480508709467648,
"calibration/batch_uniqueness": 0.887579345703125,
"calibration/buffer_distribution_entropy": 0.9004357876541927,
"calibration/buffer_entropy_100bins": 0.9118015020278871,
"calibration/buffer_entropy_10bins": 0.9004357876541927,
"calibration/buffer_entropy_50bins": 0.9146265713411662,
"calibration/confidence_entropy": 0.3413923071111721,
"calibration/coverage@0%": 0.008984375,
"calibration/coverage@1%": 0.008984375,
"calibration/coverage@10%": 0.183203125,
"calibration/coverage@15%": 0.2296875,
"calibration/coverage@20%": 0.3484375,
"calibration/coverage@25%": 0.405859375,
"calibration/coverage@30%": 0.60546875,
"calibration/coverage@5%": 0.145703125,
"calibration/distribution_entropy_10": 0.8275800902466568,
"calibration/distribution_entropy_100": 0.8529547150479395,
"calibration/ece": 0.17277675037443035,
"calibration/mean_confidence": 0.6607414673392388,
"calibration/unique_confidence_per_question": 0.9375,
"calibration/unique_confidences": 480.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 597.8,
"completions/max_terminated_length": 597.8,
"completions/mean_length": 208.18974609375,
"completions/mean_terminated_length": 208.18974609375,
"completions/min_length": 112.4,
"completions/min_terminated_length": 112.4,
"epoch": 0.288,
"grad_norm": 0.00779998954385519,
"learning_rate": 1e-06,
"loss": -0.0005,
"num_tokens": 302495197.0,
"reward": 1.112547254562378,
"reward_std": 0.1164923757314682,
"rewards/accuracy_reward": 0.54970703125,
"rewards/batchaccgated_coverage_0": 0.2503787219524384,
"rewards/batchaccgated_coverage_1": 0.2503787219524384,
"rewards/batchaccgated_coverage_10": 0.30529601573944093,
"rewards/batchaccgated_coverage_15": 0.3129884600639343,
"rewards/batchaccgated_coverage_20": 0.32123541831970215,
"rewards/batchaccgated_coverage_25": 0.3247857570648193,
"rewards/batchaccgated_coverage_5": 0.2785686820745468,
"rewards/brier_reward": 0.7823822021484375,
"rewards/confidence_uniqueness_reward": 0.9014236450195312,
"rewards/format_reward": 1.0,
"rewards/frontier_entropy_batch_reward": -0.3505002737045288,
"signal/accuracy_reward/centered_abs_mean": 0.105963134765625,
"signal/accuracy_reward/group_std_mean": 0.1416401147842407,
"signal/accuracy_reward/group_zero_std_frac": 0.590625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.7099111318588256,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0529815673828125,
"signal/advantage_abs_mean": 0.7070619702339173,
"signal/advantage_pre_scale_abs_mean": 0.08482404053211212,
"signal/advantage_pre_scale_std": 0.14781360626220702,
"signal/advantage_std": 0.9812478303909302,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.045230365544557574,
"signal/batchaccgated_coverage_0/group_std_mean": 0.0594833180308342,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.284375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.15251844823360444,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.004523036768659949,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.045230365544557574,
"signal/batchaccgated_coverage_1/group_std_mean": 0.0594833180308342,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.284375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.15251844823360444,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.004523036768659949,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.05738704428076744,
"signal/batchaccgated_coverage_10/group_std_mean": 0.07853602319955826,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.284375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.19147444814443587,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.005738704465329647,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.06193866953253746,
"signal/batchaccgated_coverage_15/group_std_mean": 0.08510458022356034,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.284375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.2074664443731308,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.006193867139518261,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.06739025488495827,
"signal/batchaccgated_coverage_20/group_std_mean": 0.09281168133020401,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.284375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.2265717178583145,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.00673902565613389,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.06950431317090988,
"signal/batchaccgated_coverage_25/group_std_mean": 0.09564992636442185,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.284375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.2345907062292099,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.006950431410223246,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.051506861299276355,
"signal/batchaccgated_coverage_5/group_std_mean": 0.06888163536787033,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.284375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.17419881820678712,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.005150686018168926,
"signal/brier_reward/centered_abs_mean": 0.11472182273864746,
"signal/brier_reward/group_std_mean": 0.15041446387767793,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.3741160035133362,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.011472182907164096,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.03848352432250977,
"signal/confidence_uniqueness_reward/group_std_mean": 0.04962219223380089,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.13226202875375748,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0038483524695038795,
"signal/format_reward/centered_abs_mean": 0.0,
"signal/format_reward/group_std_mean": 0.0,
"signal/format_reward/group_zero_std_frac": 1.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.25946560204029084,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.33681349754333495,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.06875,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.8563883185386658,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.025946559756994246,
"step": 90
},
{
"calibration/aurc": 0.24591946199026765,
"calibration/batch_distribution_entropy": 0.7075614461818885,
"calibration/batch_entropy_100bins": 0.7592024078750257,
"calibration/batch_entropy_10bins": 0.7075614461818885,
"calibration/batch_entropy_50bins": 0.7454600949405913,
"calibration/batch_uniqueness": 0.79154052734375,
"calibration/buffer_distribution_entropy": 0.8973645754390811,
"calibration/buffer_entropy_100bins": 0.9107293571887827,
"calibration/buffer_entropy_10bins": 0.8973645754390811,
"calibration/buffer_entropy_50bins": 0.9124109341610736,
"calibration/confidence_entropy": 0.276552380044214,
"calibration/coverage@0%": 0.000390625,
"calibration/coverage@1%": 0.000390625,
"calibration/coverage@10%": 0.155859375,
"calibration/coverage@15%": 0.24296875,
"calibration/coverage@20%": 0.498046875,
"calibration/coverage@25%": 0.6,
"calibration/coverage@30%": 0.7265625,
"calibration/coverage@5%": 0.033203125,
"calibration/distribution_entropy_10": 0.7075614461818885,
"calibration/distribution_entropy_100": 0.7592024078750257,
"calibration/ece": 0.17522952876068848,
"calibration/mean_confidence": 0.7383902590024226,
"calibration/unique_confidence_per_question": 0.87109375,
"calibration/unique_confidences": 446.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 673.8,
"completions/max_terminated_length": 673.8,
"completions/mean_length": 213.67978515625,
"completions/mean_terminated_length": 213.67978515625,
"completions/min_length": 106.6,
"completions/min_terminated_length": 106.6,
"epoch": 0.304,
"grad_norm": 0.0098080113530159,
"learning_rate": 1e-06,
"loss": 0.0007,
"num_tokens": 319613230.0,
"reward": 1.1096721410751342,
"reward_std": 0.11514778733253479,
"rewards/accuracy_reward": 0.53544921875,
"rewards/batchaccgated_coverage_0": 0.2749068379402161,
"rewards/batchaccgated_coverage_1": 0.2749068379402161,
"rewards/batchaccgated_coverage_10": 0.3333949983119965,
"rewards/batchaccgated_coverage_15": 0.3395045638084412,
"rewards/batchaccgated_coverage_20": 0.3442607581615448,
"rewards/batchaccgated_coverage_25": 0.3489796042442322,
"rewards/batchaccgated_coverage_5": 0.3077440559864044,
"rewards/brier_reward": 0.761984646320343,
"rewards/confidence_uniqueness_reward": 0.8424369812011718,
"rewards/format_reward": 1.0,
"rewards/frontier_entropy_batch_reward": -0.40864412784576415,
"signal/accuracy_reward/centered_abs_mean": 0.102386474609375,
"signal/accuracy_reward/group_std_mean": 0.13588072061538697,
"signal/accuracy_reward/group_zero_std_frac": 0.609375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 4.555649733543396,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0511932373046875,
"signal/advantage_abs_mean": 0.6927678704261779,
"signal/advantage_pre_scale_abs_mean": 0.08298513144254685,
"signal/advantage_pre_scale_std": 0.14709524512290956,
"signal/advantage_std": 0.9774874567985534,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.038359276950359344,
"signal/batchaccgated_coverage_0/group_std_mean": 0.05275464057922363,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.34057039618492124,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.003835927741602063,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.038359276950359344,
"signal/batchaccgated_coverage_1/group_std_mean": 0.05275464057922363,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.34057039618492124,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.003835927741602063,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.05294272229075432,
"signal/batchaccgated_coverage_10/group_std_mean": 0.07617195844650268,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.491510221362114,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0052942724898457525,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.055462636053562164,
"signal/batchaccgated_coverage_15/group_std_mean": 0.07984613552689553,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.5109684467315674,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.005546264024451375,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.058027705550193785,
"signal/batchaccgated_coverage_20/group_std_mean": 0.083106529712677,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.5336898595094681,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.00580277070403099,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.06214257776737213,
"signal/batchaccgated_coverage_25/group_std_mean": 0.08812821805477142,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.5636172443628311,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.00621425798162818,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.045113662257790565,
"signal/batchaccgated_coverage_5/group_std_mean": 0.06431550309062004,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.40672857165336607,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.004511366272345186,
"signal/brier_reward/centered_abs_mean": 0.10874119400978088,
"signal/brier_reward/group_std_mean": 0.14447961151599883,
"signal/brier_reward/group_zero_std_frac": 0.00625,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.9993231773376465,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010874119400978089,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.0689098596572876,
"signal/confidence_uniqueness_reward/group_std_mean": 0.08614575937390327,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.003125,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.6100890040397644,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.006890986301004887,
"signal/format_reward/centered_abs_mean": 0.0,
"signal/format_reward/group_std_mean": 0.0,
"signal/format_reward/group_zero_std_frac": 1.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.26197037696838377,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.33474252820014955,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.1,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 2.4359241008758543,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02619703859090805,
"step": 95
},
{
"calibration/aurc": 0.21286697073676283,
"calibration/batch_distribution_entropy": 0.7180094784510327,
"calibration/batch_entropy_100bins": 0.7695135955069387,
"calibration/batch_entropy_10bins": 0.7180094784510327,
"calibration/batch_entropy_50bins": 0.7510943831811785,
"calibration/batch_uniqueness": 0.810723876953125,
"calibration/buffer_distribution_entropy": 0.8894455245055672,
"calibration/buffer_entropy_100bins": 0.9059742188620173,
"calibration/buffer_entropy_10bins": 0.8894455245055672,
"calibration/buffer_entropy_50bins": 0.90625901533899,
"calibration/confidence_entropy": 0.2934043293048507,
"calibration/coverage@0%": 0.053125,
"calibration/coverage@1%": 0.12578125,
"calibration/coverage@10%": 0.35078125,
"calibration/coverage@15%": 0.4609375,
"calibration/coverage@20%": 0.51484375,
"calibration/coverage@25%": 0.65390625,
"calibration/coverage@30%": 0.716015625,
"calibration/coverage@5%": 0.226953125,
"calibration/distribution_entropy_10": 0.7180094784510327,
"calibration/distribution_entropy_100": 0.7695135955069387,
"calibration/ece": 0.17950482768710851,
"calibration/mean_confidence": 0.7667098239329005,
"calibration/unique_confidence_per_question": 0.859765625,
"calibration/unique_confidences": 440.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 508.0,
"completions/max_terminated_length": 508.0,
"completions/mean_length": 217.72666015625,
"completions/mean_terminated_length": 217.74772033691406,
"completions/min_length": 89.8,
"completions/min_terminated_length": 110.8,
"epoch": 0.32,
"grad_norm": 0.009086993522942066,
"learning_rate": 1e-06,
"loss": 0.0018,
"num_tokens": 336931455.0,
"reward": 1.1514826774597169,
"reward_std": 0.09998664557933808,
"rewards/accuracy_reward": 0.5525390625,
"rewards/batchaccgated_coverage_0": 0.33354930877685546,
"rewards/batchaccgated_coverage_1": 0.33354930877685546,
"rewards/batchaccgated_coverage_10": 0.38442744612693786,
"rewards/batchaccgated_coverage_15": 0.38863163590431216,
"rewards/batchaccgated_coverage_20": 0.3966011881828308,
"rewards/batchaccgated_coverage_25": 0.40016393065452577,
"rewards/batchaccgated_coverage_5": 0.36789807081222536,
"rewards/brier_reward": 0.7663132786750794,
"rewards/confidence_uniqueness_reward": 0.8366964340209961,
"rewards/format_reward": 0.99990234375,
"rewards/frontier_entropy_batch_reward": -0.45521129965782164,
"signal/accuracy_reward/centered_abs_mean": 0.07762451171875,
"signal/accuracy_reward/group_std_mean": 0.10625932067632675,
"signal/accuracy_reward/group_zero_std_frac": 0.684375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 3.14913125038147,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.038812255859375,
"signal/advantage_abs_mean": 0.6958471179008484,
"signal/advantage_pre_scale_abs_mean": 0.07106774747371673,
"signal/advantage_pre_scale_std": 0.13567830473184586,
"signal/advantage_std": 0.9776611447334289,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.030544858798384665,
"signal/batchaccgated_coverage_0/group_std_mean": 0.04142726249992847,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.23609809279441835,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0030544860754162075,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.030544858798384665,
"signal/batchaccgated_coverage_1/group_std_mean": 0.04142726249992847,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.23609809279441835,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0030544860754162075,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.04338853061199188,
"signal/batchaccgated_coverage_10/group_std_mean": 0.06283801794052124,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.3448873579502106,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0043388533405959604,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.045464062690734865,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06578328609466552,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.35565306842327116,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004546406399458647,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.05130014047026634,
"signal/batchaccgated_coverage_20/group_std_mean": 0.07374549061059951,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.3836148977279663,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.005130014196038246,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.055260340124368666,
"signal/batchaccgated_coverage_25/group_std_mean": 0.07858419045805931,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.4258087635040283,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.005526033975183964,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03807741701602936,
"signal/batchaccgated_coverage_5/group_std_mean": 0.05399175882339478,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.31254834979772567,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003807741729542613,
"signal/brier_reward/centered_abs_mean": 0.10047474503517151,
"signal/brier_reward/group_std_mean": 0.13225481063127517,
"signal/brier_reward/group_zero_std_frac": 0.003125,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.836492919921875,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010047474689781666,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.06190996468067169,
"signal/confidence_uniqueness_reward/group_std_mean": 0.0789609931409359,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.003125,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.516329950094223,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.006190996617078781,
"signal/format_reward/centered_abs_mean": 0.000189208984375,
"signal/format_reward/group_std_mean": 0.0005524271633476019,
"signal/format_reward/group_zero_std_frac": 0.996875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.004739187657833099,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 9.46044921875e-05,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2287631779909134,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2971174418926239,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.175,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.8700612783432007,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02287631928920746,
"step": 100
},
{
"epoch": 0.32,
"eval_completions/clipped_ratio": 0.0,
"eval_completions/max_length": 365.75,
"eval_completions/max_terminated_length": 365.75,
"eval_completions/mean_length": 217.65827178955078,
"eval_completions/mean_terminated_length": 217.65827178955078,
"eval_completions/min_length": 135.25,
"eval_completions/min_terminated_length": 135.25,
"eval_loss": 0.0,
"eval_num_tokens": 336931455.0,
"eval_reward": 0.7761687785387039,
"eval_reward_std": 0.25038251653313637,
"eval_rewards/accuracy_reward": 0.416015625,
"eval_rewards/batchaccgated_coverage_0": 0.02165497886016965,
"eval_rewards/batchaccgated_coverage_1": 0.02165497886016965,
"eval_rewards/batchaccgated_coverage_10": 0.021058444399386644,
"eval_rewards/batchaccgated_coverage_15": 0.020936568500474095,
"eval_rewards/batchaccgated_coverage_20": 0.01996617461554706,
"eval_rewards/batchaccgated_coverage_25": 0.018680386478081346,
"eval_rewards/batchaccgated_coverage_5": 0.02165497886016965,
"eval_rewards/brier_reward": 0.7083665430545807,
"eval_rewards/confidence_uniqueness_reward": 0.82763671875,
"eval_rewards/format_reward": 1.0,
"eval_rewards/frontier_entropy_batch_reward": -1.0,
"eval_runtime": 19.4492,
"eval_samples_per_second": 25.708,
"eval_signal/accuracy_reward/centered_abs_mean": 0.4725341796875,
"eval_signal/accuracy_reward/group_std_mean": 0.4934079945087433,
"eval_signal/accuracy_reward/group_zero_std_frac": 0.0,
"eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9441142380237579,
"eval_signal/accuracy_reward/weight": 0.5,
"eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.23626708984375,
"eval_signal/advantage_abs_mean": 0.927559033036232,
"eval_signal/advantage_pre_scale_abs_mean": 0.2325662486255169,
"eval_signal/advantage_pre_scale_std": 0.24754254519939423,
"eval_signal/advantage_std": 0.9877234101295471,
"eval_signal/batchaccgated_coverage_0/centered_abs_mean": 0.036966916639357805,
"eval_signal/batchaccgated_coverage_0/group_std_mean": 0.04998053889721632,
"eval_signal/batchaccgated_coverage_0/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.014758087694644928,
"eval_signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0036966917105019093,
"eval_signal/batchaccgated_coverage_1/centered_abs_mean": 0.036966916639357805,
"eval_signal/batchaccgated_coverage_1/group_std_mean": 0.04998053889721632,
"eval_signal/batchaccgated_coverage_1/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.014758087694644928,
"eval_signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0036966917105019093,
"eval_signal/batchaccgated_coverage_10/centered_abs_mean": 0.03580868989229202,
"eval_signal/batchaccgated_coverage_10/group_std_mean": 0.048450381495058537,
"eval_signal/batchaccgated_coverage_10/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.01430171076208353,
"eval_signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0035808690590783954,
"eval_signal/batchaccgated_coverage_15/centered_abs_mean": 0.03561611007899046,
"eval_signal/batchaccgated_coverage_15/group_std_mean": 0.04821595922112465,
"eval_signal/batchaccgated_coverage_15/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.01422567549161613,
"eval_signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0035616111708804965,
"eval_signal/batchaccgated_coverage_20/centered_abs_mean": 0.03365161083638668,
"eval_signal/batchaccgated_coverage_20/group_std_mean": 0.045826321467757225,
"eval_signal/batchaccgated_coverage_20/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.013447226490825415,
"eval_signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0033651611884124577,
"eval_signal/batchaccgated_coverage_25/centered_abs_mean": 0.030309504363685846,
"eval_signal/batchaccgated_coverage_25/group_std_mean": 0.041597504168748856,
"eval_signal/batchaccgated_coverage_25/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.01211921963840723,
"eval_signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.003030950319953263,
"eval_signal/batchaccgated_coverage_5/centered_abs_mean": 0.036966916639357805,
"eval_signal/batchaccgated_coverage_5/group_std_mean": 0.04998053889721632,
"eval_signal/batchaccgated_coverage_5/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.014758087694644928,
"eval_signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0036966917105019093,
"eval_signal/brier_reward/centered_abs_mean": 0.2990405112504959,
"eval_signal/brier_reward/group_std_mean": 0.3524217829108238,
"eval_signal/brier_reward/group_zero_std_frac": 0.0,
"eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.11960718035697937,
"eval_signal/brier_reward/weight": 0.10000000149011612,
"eval_signal/brier_reward/weighted_centered_abs_mean": 0.02990405075252056,
"eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.108489990234375,
"eval_signal/confidence_uniqueness_reward/group_std_mean": 0.12797259911894798,
"eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.04323085118085146,
"eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.010848999256268144,
"eval_signal/format_reward/centered_abs_mean": 0.0,
"eval_signal/format_reward/group_std_mean": 0.0,
"eval_signal/format_reward/group_zero_std_frac": 1.0,
"eval_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"eval_signal/format_reward/weight": 0.5,
"eval_signal/format_reward/weighted_centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"eval_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"eval_steps_per_second": 0.206,
"step": 100
},
{
"epoch": 0.32,
"step": 100,
"train_probe_completions/clipped_ratio": 0.0,
"train_probe_completions/max_length": 354.5,
"train_probe_completions/max_terminated_length": 354.5,
"train_probe_completions/mean_length": 216.45204544067383,
"train_probe_completions/mean_terminated_length": 216.45204544067383,
"train_probe_completions/min_length": 125.5,
"train_probe_completions/min_terminated_length": 125.5,
"train_probe_loss": 0.0,
"train_probe_num_tokens": 336931455.0,
"train_probe_reward": 0.8671507686376572,
"train_probe_reward_std": 0.24470600113272667,
"train_probe_rewards/accuracy_reward": 0.56640625,
"train_probe_rewards/batchaccgated_coverage_0": 0.030086406506597996,
"train_probe_rewards/batchaccgated_coverage_1": 0.030086406506597996,
"train_probe_rewards/batchaccgated_coverage_10": 0.030843764543533325,
"train_probe_rewards/batchaccgated_coverage_15": 0.03288966044783592,
"train_probe_rewards/batchaccgated_coverage_20": 0.035868712700903416,
"train_probe_rewards/batchaccgated_coverage_25": 0.04602243867702782,
"train_probe_rewards/batchaccgated_coverage_5": 0.030086406506597996,
"train_probe_rewards/brier_reward": 0.7896275818347931,
"train_probe_rewards/confidence_uniqueness_reward": 0.81396484375,
"train_probe_rewards/format_reward": 1.0,
"train_probe_rewards/frontier_entropy_batch_reward": -1.0,
"train_probe_runtime": 18.6421,
"train_probe_samples_per_second": 26.821,
"train_probe_signal/accuracy_reward/centered_abs_mean": 0.477783203125,
"train_probe_signal/accuracy_reward/group_std_mean": 0.49626488238573074,
"train_probe_signal/accuracy_reward/group_zero_std_frac": 0.0,
"train_probe_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9770863354206085,
"train_probe_signal/accuracy_reward/weight": 0.5,
"train_probe_signal/accuracy_reward/weighted_centered_abs_mean": 0.2388916015625,
"train_probe_signal/advantage_abs_mean": 0.922933354973793,
"train_probe_signal/advantage_pre_scale_abs_mean": 0.22603842988610268,
"train_probe_signal/advantage_pre_scale_std": 0.24194854870438576,
"train_probe_signal/advantage_std": 0.987713947892189,
"train_probe_signal/batchaccgated_coverage_0/centered_abs_mean": 0.06662590149790049,
"train_probe_signal/batchaccgated_coverage_0/group_std_mean": 0.09628704190254211,
"train_probe_signal/batchaccgated_coverage_0/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.02739534294232726,
"train_probe_signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0066625900799408555,
"train_probe_signal/batchaccgated_coverage_1/centered_abs_mean": 0.06662590149790049,
"train_probe_signal/batchaccgated_coverage_1/group_std_mean": 0.09628704190254211,
"train_probe_signal/batchaccgated_coverage_1/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.02739534294232726,
"train_probe_signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0066625900799408555,
"train_probe_signal/batchaccgated_coverage_10/centered_abs_mean": 0.0600652527064085,
"train_probe_signal/batchaccgated_coverage_10/group_std_mean": 0.08525818027555943,
"train_probe_signal/batchaccgated_coverage_10/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.02466637920588255,
"train_probe_signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.006006525247357786,
"train_probe_signal/batchaccgated_coverage_15/centered_abs_mean": 0.0582816731184721,
"train_probe_signal/batchaccgated_coverage_15/group_std_mean": 0.08177538216114044,
"train_probe_signal/batchaccgated_coverage_15/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.023927449248731136,
"train_probe_signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.005828167195431888,
"train_probe_signal/batchaccgated_coverage_20/centered_abs_mean": 0.051078831776976585,
"train_probe_signal/batchaccgated_coverage_20/group_std_mean": 0.07118842657655478,
"train_probe_signal/batchaccgated_coverage_20/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.020978801883757114,
"train_probe_signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.005107883189339191,
"train_probe_signal/batchaccgated_coverage_25/centered_abs_mean": 0.059190056286752224,
"train_probe_signal/batchaccgated_coverage_25/group_std_mean": 0.07565325032919645,
"train_probe_signal/batchaccgated_coverage_25/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.02438872284255922,
"train_probe_signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.005919005780015141,
"train_probe_signal/batchaccgated_coverage_5/centered_abs_mean": 0.06662590149790049,
"train_probe_signal/batchaccgated_coverage_5/group_std_mean": 0.09628704190254211,
"train_probe_signal/batchaccgated_coverage_5/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.02739534294232726,
"train_probe_signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0066625900799408555,
"train_probe_signal/brier_reward/centered_abs_mean": 0.24718650802969933,
"train_probe_signal/brier_reward/group_std_mean": 0.3167901560664177,
"train_probe_signal/brier_reward/group_zero_std_frac": 0.0,
"train_probe_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.10086004994809628,
"train_probe_signal/brier_reward/weight": 0.10000000149011612,
"train_probe_signal/brier_reward/weighted_centered_abs_mean": 0.024718651082366705,
"train_probe_signal/confidence_uniqueness_reward/centered_abs_mean": 0.1171417236328125,
"train_probe_signal/confidence_uniqueness_reward/group_std_mean": 0.13612791895866394,
"train_probe_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"train_probe_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.048000496812164783,
"train_probe_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"train_probe_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.011714172782376409,
"train_probe_signal/format_reward/centered_abs_mean": 0.0,
"train_probe_signal/format_reward/group_std_mean": 0.0,
"train_probe_signal/format_reward/group_zero_std_frac": 1.0,
"train_probe_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/format_reward/weight": 0.5,
"train_probe_signal/format_reward/weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"train_probe_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"train_probe_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"train_probe_steps_per_second": 0.215
},
{
"calibration/aurc": 0.2556721817417022,
"calibration/batch_distribution_entropy": 0.7866464019477621,
"calibration/batch_entropy_100bins": 0.817785040867891,
"calibration/batch_entropy_10bins": 0.7866464019477621,
"calibration/batch_entropy_50bins": 0.8054808227050481,
"calibration/batch_uniqueness": 0.846856689453125,
"calibration/buffer_distribution_entropy": 0.8875596698815901,
"calibration/buffer_entropy_100bins": 0.9073849890238268,
"calibration/buffer_entropy_10bins": 0.8875596698815901,
"calibration/buffer_entropy_50bins": 0.9054927837416173,
"calibration/confidence_entropy": 0.3355601176581148,
"calibration/coverage@0%": 0.008984375,
"calibration/coverage@1%": 0.008984375,
"calibration/coverage@10%": 0.151171875,
"calibration/coverage@15%": 0.256640625,
"calibration/coverage@20%": 0.31953125,
"calibration/coverage@25%": 0.527734375,
"calibration/coverage@30%": 0.65546875,
"calibration/coverage@5%": 0.057421875,
"calibration/distribution_entropy_10": 0.7866464019477621,
"calibration/distribution_entropy_100": 0.817785040867891,
"calibration/ece": 0.18152041240819028,
"calibration/mean_confidence": 0.7234203636010827,
"calibration/unique_confidence_per_question": 0.88203125,
"calibration/unique_confidences": 451.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 590.6,
"completions/max_terminated_length": 590.6,
"completions/mean_length": 216.04560546875,
"completions/mean_terminated_length": 216.06644592285156,
"completions/min_length": 87.6,
"completions/min_terminated_length": 110.8,
"epoch": 0.336,
"grad_norm": 0.011138031259179115,
"learning_rate": 1e-06,
"loss": 0.0019,
"num_tokens": 353866194.0,
"reward": 1.1516883850097657,
"reward_std": 0.10085563957691193,
"rewards/accuracy_reward": 0.5544921875,
"rewards/batchaccgated_coverage_0": 0.3259027421474457,
"rewards/batchaccgated_coverage_1": 0.3259027421474457,
"rewards/batchaccgated_coverage_10": 0.3741886556148529,
"rewards/batchaccgated_coverage_15": 0.37758533358573915,
"rewards/batchaccgated_coverage_20": 0.38451558351516724,
"rewards/batchaccgated_coverage_25": 0.38996599316596986,
"rewards/batchaccgated_coverage_5": 0.3512499928474426,
"rewards/brier_reward": 0.7832236766815186,
"rewards/confidence_uniqueness_reward": 0.8499019503593445,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.4170371055603027,
"signal/accuracy_reward/centered_abs_mean": 0.08321533203125,
"signal/accuracy_reward/group_std_mean": 0.11516006290912628,
"signal/accuracy_reward/group_zero_std_frac": 0.65625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.894821548461914,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.041607666015625,
"signal/advantage_abs_mean": 0.6907509207725525,
"signal/advantage_pre_scale_abs_mean": 0.0712737798690796,
"signal/advantage_pre_scale_std": 0.13415005505084993,
"signal/advantage_std": 0.9779654622077942,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.031348873302340506,
"signal/batchaccgated_coverage_0/group_std_mean": 0.042764414101839066,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.2213585928082466,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0031348873395472764,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.031348873302340506,
"signal/batchaccgated_coverage_1/group_std_mean": 0.042764414101839066,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.2213585928082466,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0031348873395472764,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.0436157763004303,
"signal/batchaccgated_coverage_10/group_std_mean": 0.0626350462436676,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.30962895154953,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004361577751114965,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.0447103850543499,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06406425088644027,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.3189202666282654,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004471038654446602,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.05022920444607735,
"signal/batchaccgated_coverage_20/group_std_mean": 0.07131602540612221,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.35801719427108764,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.005022920668125153,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.05553136318922043,
"signal/batchaccgated_coverage_25/group_std_mean": 0.07772955298423767,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.39893207550048826,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.005553136393427849,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.036574269831180575,
"signal/batchaccgated_coverage_5/group_std_mean": 0.05113850459456444,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.25672203302383423,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003657426917925477,
"signal/brier_reward/centered_abs_mean": 0.10104330182075501,
"signal/brier_reward/group_std_mean": 0.13122065663337706,
"signal/brier_reward/group_zero_std_frac": 0.003125,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.704465526342392,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010104330256581306,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.051836713403463366,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06646632850170135,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.40172757804393766,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.005183671461418271,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.013498052209615707,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2215854376554489,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.289985454082489,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.1375,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.5255733251571655,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.022158544138073923,
"step": 105
},
{
"calibration/aurc": 0.23596704654504355,
"calibration/batch_distribution_entropy": 0.8006939240208271,
"calibration/batch_entropy_100bins": 0.8043571210637598,
"calibration/batch_entropy_10bins": 0.8006939240208271,
"calibration/batch_entropy_50bins": 0.7900984424900207,
"calibration/batch_uniqueness": 0.8420013427734375,
"calibration/buffer_distribution_entropy": 0.8903341710152709,
"calibration/buffer_entropy_100bins": 0.9132238832491643,
"calibration/buffer_entropy_10bins": 0.8903341710152709,
"calibration/buffer_entropy_50bins": 0.9079347155328804,
"calibration/confidence_entropy": 0.2975190048757601,
"calibration/coverage@0%": 0.14296875,
"calibration/coverage@1%": 0.182421875,
"calibration/coverage@10%": 0.3640625,
"calibration/coverage@15%": 0.4265625,
"calibration/coverage@20%": 0.52421875,
"calibration/coverage@25%": 0.586328125,
"calibration/coverage@30%": 0.643359375,
"calibration/coverage@5%": 0.21640625,
"calibration/distribution_entropy_10": 0.8006939240208271,
"calibration/distribution_entropy_100": 0.8043571210637598,
"calibration/ece": 0.13748562249744056,
"calibration/mean_confidence": 0.6163752964206111,
"calibration/unique_confidence_per_question": 0.839453125,
"calibration/unique_confidences": 429.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 600.0,
"completions/max_terminated_length": 600.0,
"completions/mean_length": 211.212890625,
"completions/mean_terminated_length": 211.212890625,
"completions/min_length": 115.8,
"completions/min_terminated_length": 115.8,
"epoch": 0.352,
"grad_norm": 0.007924962788820267,
"learning_rate": 1e-06,
"loss": -0.0004,
"num_tokens": 371289430.0,
"reward": 1.080542302131653,
"reward_std": 0.10274379849433898,
"rewards/accuracy_reward": 0.48671875,
"rewards/batchaccgated_coverage_0": 0.27630739808082583,
"rewards/batchaccgated_coverage_1": 0.27630739808082583,
"rewards/batchaccgated_coverage_10": 0.3199822962284088,
"rewards/batchaccgated_coverage_15": 0.32377060055732726,
"rewards/batchaccgated_coverage_20": 0.3276528656482697,
"rewards/batchaccgated_coverage_25": 0.3298716008663177,
"rewards/batchaccgated_coverage_5": 0.30000272393226624,
"rewards/brier_reward": 0.7719616174697876,
"rewards/confidence_uniqueness_reward": 0.8616905212402344,
"rewards/format_reward": 1.0,
"rewards/frontier_entropy_batch_reward": -0.4157181799411774,
"signal/accuracy_reward/centered_abs_mean": 0.097314453125,
"signal/accuracy_reward/group_std_mean": 0.12220201343297958,
"signal/accuracy_reward/group_zero_std_frac": 0.671875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 3.603860092163086,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0486572265625,
"signal/advantage_abs_mean": 0.7189256310462951,
"signal/advantage_pre_scale_abs_mean": 0.07676824629306793,
"signal/advantage_pre_scale_std": 0.13726735413074492,
"signal/advantage_std": 0.9785060524940491,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.028924032300710677,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03871177956461906,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.35625,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.249131478369236,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0028924033045768737,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.028924032300710677,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03871177956461906,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.35625,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.249131478369236,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0028924033045768737,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.038000909984111785,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05644780248403549,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.35625,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.3176309555768967,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.003800091100856662,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03933459669351578,
"signal/batchaccgated_coverage_15/group_std_mean": 0.058353348076343535,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.35625,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.3222628504037857,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.003933459706604481,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04203208163380623,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06182496696710586,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.35625,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.33332120776176455,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004203208163380623,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04303168132901192,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06316192150115967,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.35625,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.32398168444633485,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004303168179467321,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.032129060477018356,
"signal/batchaccgated_coverage_5/group_std_mean": 0.0460076242685318,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.35625,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.28063117116689684,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0032129061408340933,
"signal/brier_reward/centered_abs_mean": 0.10276058614253998,
"signal/brier_reward/group_std_mean": 0.1344049960374832,
"signal/brier_reward/group_zero_std_frac": 0.003125,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.7991366803646087,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.01027605850249529,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.0492556095123291,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06328397393226623,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.003125,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.39417117536067964,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004925561137497425,
"signal/format_reward/centered_abs_mean": 0.0,
"signal/format_reward/group_std_mean": 0.0,
"signal/format_reward/group_zero_std_frac": 1.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.24574898183345795,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.318847519159317,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.140625,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.7696859955787658,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.024574899300932884,
"step": 110
},
{
"calibration/aurc": 0.3077015518606279,
"calibration/batch_distribution_entropy": 0.7898167862947604,
"calibration/batch_entropy_100bins": 0.8180512861057277,
"calibration/batch_entropy_10bins": 0.7898167862947604,
"calibration/batch_entropy_50bins": 0.8030328369866471,
"calibration/batch_uniqueness": 0.851416015625,
"calibration/buffer_distribution_entropy": 0.8949600693994654,
"calibration/buffer_entropy_100bins": 0.917998688853852,
"calibration/buffer_entropy_10bins": 0.8949600693994654,
"calibration/buffer_entropy_50bins": 0.9101072811232015,
"calibration/confidence_entropy": 0.31613800492907196,
"calibration/coverage@0%": 0.00390625,
"calibration/coverage@1%": 0.0875,
"calibration/coverage@10%": 0.11171875,
"calibration/coverage@15%": 0.23984375,
"calibration/coverage@20%": 0.3796875,
"calibration/coverage@25%": 0.427734375,
"calibration/coverage@30%": 0.508984375,
"calibration/coverage@5%": 0.10234375,
"calibration/distribution_entropy_10": 0.7898167862947604,
"calibration/distribution_entropy_100": 0.8180512861057277,
"calibration/ece": 0.18833717800853794,
"calibration/mean_confidence": 0.6745171109559501,
"calibration/unique_confidence_per_question": 0.84609375,
"calibration/unique_confidences": 433.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 658.2,
"completions/max_terminated_length": 658.2,
"completions/mean_length": 210.0818359375,
"completions/mean_terminated_length": 210.10162963867188,
"completions/min_length": 91.0,
"completions/min_terminated_length": 113.4,
"epoch": 0.368,
"grad_norm": 0.008605880662798882,
"learning_rate": 1e-06,
"loss": 0.0007,
"num_tokens": 388506140.0,
"reward": 1.1214004278182983,
"reward_std": 0.09413044750690461,
"rewards/accuracy_reward": 0.517578125,
"rewards/batchaccgated_coverage_0": 0.296656060218811,
"rewards/batchaccgated_coverage_1": 0.296656060218811,
"rewards/batchaccgated_coverage_10": 0.34905418157577517,
"rewards/batchaccgated_coverage_15": 0.35381984114646914,
"rewards/batchaccgated_coverage_20": 0.35733194947242736,
"rewards/batchaccgated_coverage_25": 0.3588412344455719,
"rewards/batchaccgated_coverage_5": 0.33130133152008057,
"rewards/brier_reward": 0.7963966488838196,
"rewards/confidence_uniqueness_reward": 0.8733832955360412,
"rewards/format_reward": 0.99990234375,
"rewards/frontier_entropy_batch_reward": -0.38683871030807493,
"signal/accuracy_reward/centered_abs_mean": 0.07576904296875,
"signal/accuracy_reward/group_std_mean": 0.10405463427305221,
"signal/accuracy_reward/group_zero_std_frac": 0.684375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 5.112514925003052,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.037884521484375,
"signal/advantage_abs_mean": 0.6992959856987,
"signal/advantage_pre_scale_abs_mean": 0.06689565181732178,
"signal/advantage_pre_scale_std": 0.12708753049373628,
"signal/advantage_std": 0.9756571412086487,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.031997333467006686,
"signal/batchaccgated_coverage_0/group_std_mean": 0.042501792311668396,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.34375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.43886796236038206,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0031997334212064744,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.031997333467006686,
"signal/batchaccgated_coverage_1/group_std_mean": 0.042501792311668396,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.34375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.43886796236038206,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0031997334212064744,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.0443273775279522,
"signal/batchaccgated_coverage_10/group_std_mean": 0.06273244768381118,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.34375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.6147834897041321,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004432737827301025,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.04662763997912407,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06607433557510375,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.34375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.6436557769775391,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004662763979285956,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04926467537879944,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06952771916985512,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.34375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.6920301258563996,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.00492646787315607,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.05098762735724449,
"signal/batchaccgated_coverage_25/group_std_mean": 0.07162266075611115,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.34375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.7069201648235321,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.005098763015121221,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03900737836956978,
"signal/batchaccgated_coverage_5/group_std_mean": 0.053840626031160355,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.34375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.5482029139995575,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0039007378742098807,
"signal/brier_reward/centered_abs_mean": 0.08583471775054932,
"signal/brier_reward/group_std_mean": 0.11394892483949662,
"signal/brier_reward/group_zero_std_frac": 0.00625,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 1.2062372267246246,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008583471924066544,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.04261606186628342,
"signal/confidence_uniqueness_reward/group_std_mean": 0.05452136397361755,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.009375,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.532250988483429,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004261606372892857,
"signal/format_reward/centered_abs_mean": 0.000189208984375,
"signal/format_reward/group_std_mean": 0.0005524271633476019,
"signal/format_reward/group_zero_std_frac": 0.996875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.006750371307134628,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 9.46044921875e-05,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.22840916812419892,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2947576820850372,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.128125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 3.2053250074386597,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.022840917855501176,
"step": 115
},
{
"calibration/aurc": 0.2877203310665097,
"calibration/batch_distribution_entropy": 0.8508872195773087,
"calibration/batch_entropy_100bins": 0.8596648112531013,
"calibration/batch_entropy_10bins": 0.8508872195773087,
"calibration/batch_entropy_50bins": 0.8514976665670577,
"calibration/batch_uniqueness": 0.8954345703125,
"calibration/buffer_distribution_entropy": 0.8946714253710242,
"calibration/buffer_entropy_100bins": 0.9180899094919133,
"calibration/buffer_entropy_10bins": 0.8946714253710242,
"calibration/buffer_entropy_50bins": 0.9084049182184752,
"calibration/confidence_entropy": 0.35298891835525925,
"calibration/coverage@0%": 0.153515625,
"calibration/coverage@1%": 0.16640625,
"calibration/coverage@10%": 0.29140625,
"calibration/coverage@15%": 0.330078125,
"calibration/coverage@20%": 0.3671875,
"calibration/coverage@25%": 0.428125,
"calibration/coverage@30%": 0.50703125,
"calibration/coverage@5%": 0.25859375,
"calibration/distribution_entropy_10": 0.8508872195773087,
"calibration/distribution_entropy_100": 0.8596648112531013,
"calibration/ece": 0.17890031924512512,
"calibration/mean_confidence": 0.6212225282319486,
"calibration/unique_confidence_per_question": 0.87578125,
"calibration/unique_confidences": 448.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 712.6,
"completions/max_terminated_length": 712.6,
"completions/mean_length": 210.25224609375,
"completions/mean_terminated_length": 210.2943328857422,
"completions/min_length": 65.8,
"completions/min_terminated_length": 106.4,
"epoch": 0.384,
"grad_norm": 0.0101792486384511,
"learning_rate": 1e-06,
"loss": -0.0016,
"num_tokens": 405515635.0,
"reward": 1.1468703746795654,
"reward_std": 0.09761902093887329,
"rewards/accuracy_reward": 0.5435546875,
"rewards/batchaccgated_coverage_0": 0.31184246242046354,
"rewards/batchaccgated_coverage_1": 0.31184246242046354,
"rewards/batchaccgated_coverage_10": 0.37447665333747865,
"rewards/batchaccgated_coverage_15": 0.3783999264240265,
"rewards/batchaccgated_coverage_20": 0.3803004801273346,
"rewards/batchaccgated_coverage_25": 0.3822021961212158,
"rewards/batchaccgated_coverage_5": 0.35037162005901334,
"rewards/brier_reward": 0.8052679657936096,
"rewards/confidence_uniqueness_reward": 0.8585785269737244,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.40137566328048707,
"signal/accuracy_reward/centered_abs_mean": 0.08345947265625,
"signal/accuracy_reward/group_std_mean": 0.11401332467794419,
"signal/accuracy_reward/group_zero_std_frac": 0.653125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 3.7995605945587156,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.041729736328125,
"signal/advantage_abs_mean": 0.68350191116333,
"signal/advantage_pre_scale_abs_mean": 0.06831767559051513,
"signal/advantage_pre_scale_std": 0.13085786551237105,
"signal/advantage_std": 0.9757751941680908,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.028213725239038468,
"signal/batchaccgated_coverage_0/group_std_mean": 0.0376923643052578,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.2931238904595375,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.002821372449398041,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.028213725239038468,
"signal/batchaccgated_coverage_1/group_std_mean": 0.0376923643052578,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.2931238904595375,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.002821372449398041,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.04215507581830025,
"signal/batchaccgated_coverage_10/group_std_mean": 0.06263689845800399,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.4335298717021942,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004215507628396153,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.04399342015385628,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06547538340091705,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.44530953764915465,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004399342136457563,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04558649361133575,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06763089522719383,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.47427464127540586,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004558649566024542,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04617611840367317,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06849604845046997,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.48964144885540006,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004617611691355705,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03511584736406803,
"signal/batchaccgated_coverage_5/group_std_mean": 0.05061802640557289,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.3738081336021423,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003511584876105189,
"signal/brier_reward/centered_abs_mean": 0.08375445753335953,
"signal/brier_reward/group_std_mean": 0.11226186901330948,
"signal/brier_reward/group_zero_std_frac": 0.00625,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.7786072731018067,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008375446125864983,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.04671300426125526,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06126299202442169,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.47258659899234773,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004671300668269396,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.017767124623060227,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2251691311597824,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2954529881477356,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.128125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 2.142231583595276,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0225169125944376,
"step": 120
},
{
"calibration/aurc": 0.3607472263802859,
"calibration/batch_distribution_entropy": 0.7500122341970511,
"calibration/batch_entropy_100bins": 0.8049542400220764,
"calibration/batch_entropy_10bins": 0.7500122341970511,
"calibration/batch_entropy_50bins": 0.787167590017182,
"calibration/batch_uniqueness": 0.8521942138671875,
"calibration/buffer_distribution_entropy": 0.8887323148908756,
"calibration/buffer_entropy_100bins": 0.9137444888144204,
"calibration/buffer_entropy_10bins": 0.8887323148908756,
"calibration/buffer_entropy_50bins": 0.9029990473361961,
"calibration/confidence_entropy": 0.3103455913941388,
"calibration/coverage@0%": 0.005078125,
"calibration/coverage@1%": 0.005078125,
"calibration/coverage@10%": 0.048046875,
"calibration/coverage@15%": 0.09140625,
"calibration/coverage@20%": 0.17734375,
"calibration/coverage@25%": 0.237109375,
"calibration/coverage@30%": 0.398046875,
"calibration/coverage@5%": 0.005078125,
"calibration/distribution_entropy_10": 0.7500122341970511,
"calibration/distribution_entropy_100": 0.8049542400220764,
"calibration/ece": 0.2932147729587164,
"calibration/mean_confidence": 0.7528071757315636,
"calibration/unique_confidence_per_question": 0.862109375,
"calibration/unique_confidences": 441.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 541.2,
"completions/max_terminated_length": 541.2,
"completions/mean_length": 210.08193359375,
"completions/mean_terminated_length": 210.08193359375,
"completions/min_length": 109.6,
"completions/min_terminated_length": 109.6,
"epoch": 0.4,
"grad_norm": 0.010229752399027348,
"learning_rate": 1e-06,
"loss": 0.0014,
"num_tokens": 422703322.0,
"reward": 1.1232687950134277,
"reward_std": 0.1085848331451416,
"rewards/accuracy_reward": 0.53310546875,
"rewards/batchaccgated_coverage_0": 0.30715653896331785,
"rewards/batchaccgated_coverage_1": 0.30715653896331785,
"rewards/batchaccgated_coverage_10": 0.37526739239692686,
"rewards/batchaccgated_coverage_15": 0.37950422167778014,
"rewards/batchaccgated_coverage_20": 0.3865469455718994,
"rewards/batchaccgated_coverage_25": 0.38814383149147036,
"rewards/batchaccgated_coverage_5": 0.3556707799434662,
"rewards/brier_reward": 0.7472327589988709,
"rewards/confidence_uniqueness_reward": 0.7995094299316406,
"rewards/format_reward": 1.0,
"rewards/frontier_entropy_batch_reward": -0.479028058052063,
"signal/accuracy_reward/centered_abs_mean": 0.097686767578125,
"signal/accuracy_reward/group_std_mean": 0.13303740918636323,
"signal/accuracy_reward/group_zero_std_frac": 0.6,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 9.457622385025024,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0488433837890625,
"signal/advantage_abs_mean": 0.6940873980522155,
"signal/advantage_pre_scale_abs_mean": 0.07784865200519561,
"signal/advantage_pre_scale_std": 0.14691626131534577,
"signal/advantage_std": 0.9719500541687012,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.021109426394104958,
"signal/batchaccgated_coverage_0/group_std_mean": 0.029066773131489754,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.41222373843193055,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0021109426161274314,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.021109426394104958,
"signal/batchaccgated_coverage_1/group_std_mean": 0.029066773131489754,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.41222373843193055,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0021109426161274314,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03453552387654781,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05405896380543709,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.6234773635864258,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0034535524901002645,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.037094959244132045,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05753937140107155,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.6831642776727677,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0037094960454851387,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04237879179418087,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06447049230337143,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.8257112801074982,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004237879207357765,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.044358334690332415,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06675799861550331,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.8595310986042023,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004435833543539047,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.028129175305366516,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04339451566338539,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.5461988747119904,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0028129175771027803,
"signal/brier_reward/centered_abs_mean": 0.10875137150287628,
"signal/brier_reward/group_std_mean": 0.14101923257112503,
"signal/brier_reward/group_zero_std_frac": 0.01875,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 2.253282290697098,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010875137615948915,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.06688582897186279,
"signal/confidence_uniqueness_reward/group_std_mean": 0.0849440835416317,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0125,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 1.318751060962677,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.006688582990318537,
"signal/format_reward/centered_abs_mean": 0.0,
"signal/format_reward/group_std_mean": 0.0,
"signal/format_reward/group_zero_std_frac": 1.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.20507954061031342,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.26689510345458983,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.225,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 4.196206450462341,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02050795517861843,
"step": 125
},
{
"calibration/aurc": 0.2776229963553084,
"calibration/batch_distribution_entropy": 0.7423036688398358,
"calibration/batch_entropy_100bins": 0.7839451977440045,
"calibration/batch_entropy_10bins": 0.7423036688398358,
"calibration/batch_entropy_50bins": 0.7688862412328662,
"calibration/batch_uniqueness": 0.8245369487265265,
"calibration/buffer_distribution_entropy": 0.8706244826814921,
"calibration/buffer_entropy_100bins": 0.9021279746659363,
"calibration/buffer_entropy_10bins": 0.8706244826814921,
"calibration/buffer_entropy_50bins": 0.8891632647946942,
"calibration/confidence_entropy": 0.2903092468990358,
"calibration/coverage@0%": 0.0035294117647058825,
"calibration/coverage@1%": 0.0035294117647058825,
"calibration/coverage@10%": 0.1495481004901961,
"calibration/coverage@15%": 0.1816528799019608,
"calibration/coverage@20%": 0.20318167892156863,
"calibration/coverage@25%": 0.5063449754901961,
"calibration/coverage@30%": 0.6153906250000001,
"calibration/coverage@5%": 0.10452205882352941,
"calibration/distribution_entropy_10": 0.7423036688398358,
"calibration/distribution_entropy_100": 0.7839451977440045,
"calibration/ece": 0.2396569428713164,
"calibration/mean_confidence": 0.7307854842540301,
"calibration/unique_confidence_per_question": 0.8484375,
"calibration/unique_confidences": 434.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00029296875,
"completions/max_length": 598.8,
"completions/max_terminated_length": 598.8,
"completions/mean_length": 209.8515625,
"completions/mean_terminated_length": 209.9148376464844,
"completions/min_length": 66.8,
"completions/min_terminated_length": 112.6,
"epoch": 0.416,
"grad_norm": 0.0075605930760502815,
"learning_rate": 1e-06,
"loss": 0.0013,
"num_tokens": 439733386.0,
"reward": 1.1124870777130127,
"reward_std": 0.10064191967248917,
"rewards/accuracy_reward": 0.5189453125,
"rewards/batchaccgated_coverage_0": 0.29120186567306516,
"rewards/batchaccgated_coverage_1": 0.29120186567306516,
"rewards/batchaccgated_coverage_10": 0.35079731345176696,
"rewards/batchaccgated_coverage_15": 0.35612263679504397,
"rewards/batchaccgated_coverage_20": 0.360295170545578,
"rewards/batchaccgated_coverage_25": 0.36328245997428893,
"rewards/batchaccgated_coverage_5": 0.3226629585027695,
"rewards/brier_reward": 0.7726193904876709,
"rewards/confidence_uniqueness_reward": 0.8489565372467041,
"rewards/format_reward": 0.999609375,
"rewards/frontier_entropy_batch_reward": -0.42504324316978453,
"signal/accuracy_reward/centered_abs_mean": 0.09075927734375,
"signal/accuracy_reward/group_std_mean": 0.12097106128931046,
"signal/accuracy_reward/group_zero_std_frac": 0.653125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 6.237020778656006,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.045379638671875,
"signal/advantage_abs_mean": 0.7071460604667663,
"signal/advantage_pre_scale_abs_mean": 0.07220878303050995,
"signal/advantage_pre_scale_std": 0.13626499176025392,
"signal/advantage_std": 0.9739294290542603,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.024955723620951175,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03453271798789501,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.3557733744382858,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0024955724366009235,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.024955723620951175,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03453271798789501,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.3557733744382858,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0024955724366009235,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03764743208885193,
"signal/batchaccgated_coverage_10/group_std_mean": 0.056462712585926056,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.5319705545902252,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0037647434044629337,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.04012564718723297,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05981220975518227,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.5909531474113464,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004012564849108458,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.0415158286690712,
"signal/batchaccgated_coverage_20/group_std_mean": 0.061671043187379836,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.6135587334632874,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004151582717895508,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04519175440073013,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06589726135134696,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.711653470993042,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004519175365567207,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.029009561240673064,
"signal/batchaccgated_coverage_5/group_std_mean": 0.042752762138843534,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.4274468198418617,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0029009563382714986,
"signal/brier_reward/centered_abs_mean": 0.10814433693885803,
"signal/brier_reward/group_std_mean": 0.13972438275814056,
"signal/brier_reward/group_zero_std_frac": 0.009375,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 1.5724826812744142,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010814433917403221,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.048478265106678006,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06364531964063644,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.00625,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.7787885367870331,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004847826529294252,
"signal/format_reward/centered_abs_mean": 0.0007568359375,
"signal/format_reward/group_std_mean": 0.0022097086068242787,
"signal/format_reward/group_zero_std_frac": 0.9875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.024576152861118316,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.00037841796875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.21063787639141082,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2802739828824997,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.1875,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 3.079525852203369,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02106378823518753,
"step": 130
},
{
"calibration/aurc": 0.2229856185109758,
"calibration/batch_distribution_entropy": 0.7526133420933918,
"calibration/batch_entropy_100bins": 0.7692700502971216,
"calibration/batch_entropy_10bins": 0.7526133420933918,
"calibration/batch_entropy_50bins": 0.7453037832745792,
"calibration/batch_uniqueness": 0.8147974249239756,
"calibration/buffer_distribution_entropy": 0.8545848053549738,
"calibration/buffer_entropy_100bins": 0.8908519561493037,
"calibration/buffer_entropy_10bins": 0.8545848053549738,
"calibration/buffer_entropy_50bins": 0.8758029204041107,
"calibration/confidence_entropy": 0.2689103662104882,
"calibration/coverage@0%": 0.09966770014734774,
"calibration/coverage@1%": 0.10281111861493124,
"calibration/coverage@10%": 0.28139120825147346,
"calibration/coverage@15%": 0.3459502394400786,
"calibration/coverage@20%": 0.48512325024557956,
"calibration/coverage@25%": 0.5820788310412575,
"calibration/coverage@30%": 0.6555946095284872,
"calibration/coverage@5%": 0.17510206900785855,
"calibration/distribution_entropy_10": 0.7526133420933918,
"calibration/distribution_entropy_100": 0.7692700502971216,
"calibration/ece": 0.17261090854717565,
"calibration/mean_confidence": 0.675311378139298,
"calibration/unique_confidence_per_question": 0.829296875,
"calibration/unique_confidences": 424.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00029296875,
"completions/max_length": 513.2,
"completions/max_terminated_length": 513.2,
"completions/mean_length": 198.70830078125,
"completions/mean_terminated_length": 198.76768188476564,
"completions/min_length": 82.4,
"completions/min_terminated_length": 102.4,
"epoch": 0.432,
"grad_norm": 0.013178535737097263,
"learning_rate": 1e-06,
"loss": -0.0007,
"num_tokens": 456782495.0,
"reward": 1.1549261569976808,
"reward_std": 0.09908276796340942,
"rewards/accuracy_reward": 0.5490234375,
"rewards/batchaccgated_coverage_0": 0.3268318921327591,
"rewards/batchaccgated_coverage_1": 0.3268318921327591,
"rewards/batchaccgated_coverage_10": 0.3679559826850891,
"rewards/batchaccgated_coverage_15": 0.3733347475528717,
"rewards/batchaccgated_coverage_20": 0.3791102707386017,
"rewards/batchaccgated_coverage_25": 0.3816304385662079,
"rewards/batchaccgated_coverage_5": 0.3541822642087936,
"rewards/brier_reward": 0.8274408459663392,
"rewards/confidence_uniqueness_reward": 0.8565916657447815,
"rewards/format_reward": 0.99970703125,
"rewards/frontier_entropy_batch_reward": -0.3883012652397156,
"signal/accuracy_reward/centered_abs_mean": 0.0912841796875,
"signal/accuracy_reward/group_std_mean": 0.1219587504863739,
"signal/accuracy_reward/group_zero_std_frac": 0.640625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 5.192121315002441,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.04564208984375,
"signal/advantage_abs_mean": 0.6994383931159973,
"signal/advantage_pre_scale_abs_mean": 0.07214328348636627,
"signal/advantage_pre_scale_std": 0.1338247686624527,
"signal/advantage_std": 0.9752950072288513,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.030544081702828407,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03971017673611641,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.275,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.3140122652053833,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.003054408123716712,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.030544081702828407,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03971017673611641,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.275,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.3140122652053833,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.003054408123716712,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.0391458734869957,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05446720570325851,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.275,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.4162097930908203,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.003914587432518601,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.04218263924121857,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05873872861266136,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.275,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.45109961926937103,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004218264017254114,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04443009942770004,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06213266327977181,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.275,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.4902482807636261,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004443010035902262,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.046240387856960295,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06440587341785431,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.275,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.5161166340112686,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.0046240388415753845,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.035644286498427394,
"signal/batchaccgated_coverage_5/group_std_mean": 0.048659203201532365,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.275,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.38340542912483216,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003564428770914674,
"signal/brier_reward/centered_abs_mean": 0.08845999389886856,
"signal/brier_reward/group_std_mean": 0.1160064995288849,
"signal/brier_reward/group_zero_std_frac": 0.00625,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 1.0444548368453979,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008845999930053949,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.04503092765808105,
"signal/confidence_uniqueness_reward/group_std_mean": 0.058131980895996097,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.4736461818218231,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.00450309282168746,
"signal/format_reward/centered_abs_mean": 0.000531005859375,
"signal/format_reward/group_std_mean": 0.000925451796501875,
"signal/format_reward/group_zero_std_frac": 0.996875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.014521975815296174,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0002655029296875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2255818724632263,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.30061933398246765,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.134375,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 2.66522855758667,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.022558188438415526,
"step": 135
},
{
"calibration/aurc": 0.22680256574322474,
"calibration/batch_distribution_entropy": 0.7586393963707628,
"calibration/batch_entropy_100bins": 0.811647882494239,
"calibration/batch_entropy_10bins": 0.7586393963707628,
"calibration/batch_entropy_50bins": 0.7962729484157975,
"calibration/batch_uniqueness": 0.8580333968309389,
"calibration/buffer_distribution_entropy": 0.840451071682903,
"calibration/buffer_entropy_100bins": 0.8794139136748902,
"calibration/buffer_entropy_10bins": 0.840451071682903,
"calibration/buffer_entropy_50bins": 0.8624010522687519,
"calibration/confidence_entropy": 0.3241447099529916,
"calibration/coverage@0%": 0.033203125,
"calibration/coverage@1%": 0.0390625,
"calibration/coverage@10%": 0.15,
"calibration/coverage@15%": 0.296875,
"calibration/coverage@20%": 0.45390625,
"calibration/coverage@25%": 0.6647008011252445,
"calibration/coverage@30%": 0.7491186093444228,
"calibration/coverage@5%": 0.050390625,
"calibration/distribution_entropy_10": 0.7586393963707628,
"calibration/distribution_entropy_100": 0.811647882494239,
"calibration/ece": 0.20984292933854803,
"calibration/mean_confidence": 0.7537784488416881,
"calibration/unique_confidence_per_question": 0.85390625,
"calibration/unique_confidences": 437.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 512.2,
"completions/max_terminated_length": 512.2,
"completions/mean_length": 197.61435546875,
"completions/mean_terminated_length": 197.65279541015624,
"completions/min_length": 64.4,
"completions/min_terminated_length": 105.8,
"epoch": 0.448,
"grad_norm": 0.007255893666297197,
"learning_rate": 1e-06,
"loss": -0.0021,
"num_tokens": 473758866.0,
"reward": 1.129861068725586,
"reward_std": 0.10040710270404815,
"rewards/accuracy_reward": 0.528125,
"rewards/batchaccgated_coverage_0": 0.29921073019504546,
"rewards/batchaccgated_coverage_1": 0.29921073019504546,
"rewards/batchaccgated_coverage_10": 0.35485957860946654,
"rewards/batchaccgated_coverage_15": 0.3585952401161194,
"rewards/batchaccgated_coverage_20": 0.3654420018196106,
"rewards/batchaccgated_coverage_25": 0.3677010715007782,
"rewards/batchaccgated_coverage_5": 0.33968154191970823,
"rewards/brier_reward": 0.7981340885162354,
"rewards/confidence_uniqueness_reward": 0.8771803736686706,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.4010528862476349,
"signal/accuracy_reward/centered_abs_mean": 0.09295654296875,
"signal/accuracy_reward/group_std_mean": 0.1208529606461525,
"signal/accuracy_reward/group_zero_std_frac": 0.653125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.7826510667800903,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.046478271484375,
"signal/advantage_abs_mean": 0.7077152252197265,
"signal/advantage_pre_scale_abs_mean": 0.07367970943450927,
"signal/advantage_pre_scale_std": 0.13426807075738906,
"signal/advantage_std": 0.9788227915763855,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.02387251555919647,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03293117135763168,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.13176878541707993,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0023872516117990016,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.02387251555919647,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03293117135763168,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.13176878541707993,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0023872516117990016,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03575364984571934,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05332083478569984,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.325,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.2026536673307419,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0035753651056438683,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.037353414297103885,
"signal/batchaccgated_coverage_15/group_std_mean": 0.0556136317551136,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.325,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.21399764269590377,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0037353414110839365,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04187020063400269,
"signal/batchaccgated_coverage_20/group_std_mean": 0.061674417555332185,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.325,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.24235419929027557,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004187020100653171,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.043998292088508605,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06423453614115715,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.325,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.2552462860941887,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004399829264730215,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.030366695299744607,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04537146240472793,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.325,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.16626114249229432,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0030366695718839765,
"signal/brier_reward/centered_abs_mean": 0.09086526930332184,
"signal/brier_reward/group_std_mean": 0.1190246433019638,
"signal/brier_reward/group_zero_std_frac": 0.003125,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.5565635651350022,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009086526930332184,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.04014613032341004,
"signal/confidence_uniqueness_reward/group_std_mean": 0.05144088342785835,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.003125,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.2941544473171234,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004014613013714552,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.010549094527959824,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.22650024592876433,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.29723013043403623,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.331903338432312,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02265002503991127,
"step": 140
},
{
"calibration/aurc": 0.3190162293520607,
"calibration/batch_distribution_entropy": 0.7733865423027384,
"calibration/batch_entropy_100bins": 0.8258252264505893,
"calibration/batch_entropy_10bins": 0.7733865423027384,
"calibration/batch_entropy_50bins": 0.812749547416594,
"calibration/batch_uniqueness": 0.8723682445681986,
"calibration/buffer_distribution_entropy": 0.830045858550623,
"calibration/buffer_entropy_100bins": 0.8730166426934025,
"calibration/buffer_entropy_10bins": 0.830045858550623,
"calibration/buffer_entropy_50bins": 0.8547863246401788,
"calibration/confidence_entropy": 0.3228498247152701,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.024657534246575342,
"calibration/coverage@15%": 0.1160821306262231,
"calibration/coverage@20%": 0.26946321550880625,
"calibration/coverage@25%": 0.3796485139432485,
"calibration/coverage@30%": 0.5324088796477495,
"calibration/coverage@5%": 0.018395303326810174,
"calibration/distribution_entropy_10": 0.7733865423027384,
"calibration/distribution_entropy_100": 0.8258252264505893,
"calibration/ece": 0.23615844400740132,
"calibration/mean_confidence": 0.7183754388110835,
"calibration/unique_confidence_per_question": 0.87265625,
"calibration/unique_confidences": 446.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 518.6,
"completions/max_terminated_length": 518.6,
"completions/mean_length": 207.97197265625,
"completions/mean_terminated_length": 208.01221618652343,
"completions/min_length": 66.6,
"completions/min_terminated_length": 109.4,
"epoch": 0.464,
"grad_norm": 0.01686267741024494,
"learning_rate": 1e-06,
"loss": -0.0034,
"num_tokens": 491059315.0,
"reward": 1.0841882228851318,
"reward_std": 0.09508764147758483,
"rewards/accuracy_reward": 0.47626953125,
"rewards/batchaccgated_coverage_0": 0.28371925950050353,
"rewards/batchaccgated_coverage_1": 0.28371925950050353,
"rewards/batchaccgated_coverage_10": 0.3427785158157349,
"rewards/batchaccgated_coverage_15": 0.34634679555892944,
"rewards/batchaccgated_coverage_20": 0.35073259472846985,
"rewards/batchaccgated_coverage_25": 0.35143186450004577,
"rewards/batchaccgated_coverage_5": 0.3243906080722809,
"rewards/brier_reward": 0.7432479739189148,
"rewards/confidence_uniqueness_reward": 0.8515275359153748,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.4163830876350403,
"signal/accuracy_reward/centered_abs_mean": 0.068646240234375,
"signal/accuracy_reward/group_std_mean": 0.09812315553426743,
"signal/accuracy_reward/group_zero_std_frac": 0.6875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.691566824913025,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0343231201171875,
"signal/advantage_abs_mean": 0.6976139068603515,
"signal/advantage_pre_scale_abs_mean": 0.06636083349585534,
"signal/advantage_pre_scale_std": 0.12976469546556474,
"signal/advantage_std": 0.9769352436065674,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.023193218745291233,
"signal/batchaccgated_coverage_0/group_std_mean": 0.031878309696912764,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.19712574928998947,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0023193219676613808,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.023193218745291233,
"signal/batchaccgated_coverage_1/group_std_mean": 0.031878309696912764,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.19712574928998947,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0023193219676613808,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03632913082838059,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05431736335158348,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.384375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.33260212391614913,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.003632913064211607,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.038447539508342746,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05705316588282585,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.384375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.35590308904647827,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.003844754071906209,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.041587001830339435,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06121674850583077,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.384375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.39032456278800964,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0041587001644074915,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.042377833276987076,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06213424354791641,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.384375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.3935665011405945,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004237783374264836,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.030533018335700036,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04521164000034332,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.26283711940050125,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0030533017590641974,
"signal/brier_reward/centered_abs_mean": 0.10415265560150147,
"signal/brier_reward/group_std_mean": 0.13536078333854676,
"signal/brier_reward/group_zero_std_frac": 0.003125,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.9842695713043212,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010415265895426273,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.05231492221355438,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06736963838338852,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.00625,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.5877267986536026,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.005231492361053825,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.008772870525717735,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.22115446627140045,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.29072344303131104,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.15625,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.9862762928009032,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.022115445882081985,
"step": 145
},
{
"calibration/aurc": 0.21224049764874184,
"calibration/batch_distribution_entropy": 0.7798415322938187,
"calibration/batch_entropy_100bins": 0.8147911642584983,
"calibration/batch_entropy_10bins": 0.7798415322938187,
"calibration/batch_entropy_50bins": 0.7961625161645828,
"calibration/batch_uniqueness": 0.8668152455369672,
"calibration/buffer_distribution_entropy": 0.8209975703736856,
"calibration/buffer_entropy_100bins": 0.8653531989222945,
"calibration/buffer_entropy_10bins": 0.8209975703736856,
"calibration/buffer_entropy_50bins": 0.8457953765766598,
"calibration/confidence_entropy": 0.30694821863561106,
"calibration/coverage@0%": 0.058203125,
"calibration/coverage@1%": 0.062890625,
"calibration/coverage@10%": 0.19186949608610568,
"calibration/coverage@15%": 0.45943003913894326,
"calibration/coverage@20%": 0.5333292563600783,
"calibration/coverage@25%": 0.6501934014187867,
"calibration/coverage@30%": 0.7342060604207437,
"calibration/coverage@5%": 0.11796875,
"calibration/distribution_entropy_10": 0.7798415322938187,
"calibration/distribution_entropy_100": 0.8147911642584983,
"calibration/ece": 0.15468002773798323,
"calibration/mean_confidence": 0.6992438845404393,
"calibration/unique_confidence_per_question": 0.840234375,
"calibration/unique_confidences": 430.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0005859375,
"completions/max_length": 659.6,
"completions/max_terminated_length": 659.6,
"completions/mean_length": 214.44619140625,
"completions/mean_terminated_length": 214.57216186523436,
"completions/min_length": 0.0,
"completions/min_terminated_length": 112.0,
"epoch": 0.48,
"grad_norm": 0.0105704665184021,
"learning_rate": 1e-06,
"loss": 0.002,
"num_tokens": 508303276.0,
"reward": 1.1211968660354614,
"reward_std": 0.10856495946645736,
"rewards/accuracy_reward": 0.5283203125,
"rewards/batchaccgated_coverage_0": 0.3046718329191208,
"rewards/batchaccgated_coverage_1": 0.3046718329191208,
"rewards/batchaccgated_coverage_10": 0.3373963415622711,
"rewards/batchaccgated_coverage_15": 0.34402522444725037,
"rewards/batchaccgated_coverage_20": 0.3507494151592255,
"rewards/batchaccgated_coverage_25": 0.3525814116001129,
"rewards/batchaccgated_coverage_5": 0.3194428324699402,
"rewards/brier_reward": 0.7929817199707031,
"rewards/confidence_uniqueness_reward": 0.8774391055107117,
"rewards/format_reward": 0.99931640625,
"rewards/frontier_entropy_batch_reward": -0.410174697637558,
"signal/accuracy_reward/centered_abs_mean": 0.1084228515625,
"signal/accuracy_reward/group_std_mean": 0.1389655962586403,
"signal/accuracy_reward/group_zero_std_frac": 0.61875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 5.808750772476197,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.05421142578125,
"signal/advantage_abs_mean": 0.7242170333862304,
"signal/advantage_pre_scale_abs_mean": 0.08107282370328903,
"signal/advantage_pre_scale_std": 0.14269192516803741,
"signal/advantage_std": 0.976265835762024,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.03433732762932777,
"signal/batchaccgated_coverage_0/group_std_mean": 0.045710965245962146,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.35650462210178374,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0034337328746914864,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.03433732762932777,
"signal/batchaccgated_coverage_1/group_std_mean": 0.045710965245962146,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.35650462210178374,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0034337328746914864,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.041103479266166684,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05710596069693565,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.4215922772884369,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004110347898676992,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.043928731232881546,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06079170107841492,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.44510679244995116,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004392873216420412,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.048879412561655046,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06712483316659927,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.48369852602481844,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004887941107153892,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.05033530816435814,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06883393973112106,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.49228758811950685,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.005033530574291944,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.0370772659778595,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04988641068339348,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.3834536552429199,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003707726625725627,
"signal/brier_reward/centered_abs_mean": 0.10277245491743088,
"signal/brier_reward/group_std_mean": 0.13592770397663118,
"signal/brier_reward/group_zero_std_frac": 0.003125,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 1.0663122713565827,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010277245379984378,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.04252248294651508,
"signal/confidence_uniqueness_reward/group_std_mean": 0.05559176430106163,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.003125,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.4616075992584229,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004252248397096992,
"signal/format_reward/centered_abs_mean": 0.001312255859375,
"signal/format_reward/group_std_mean": 0.0035306816920638085,
"signal/format_reward/group_zero_std_frac": 0.98125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.07763538397848606,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0006561279296875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.23732090890407562,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.30878318548202516,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.15,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 2.5463720083236696,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02373209185898304,
"step": 150
},
{
"epoch": 0.48,
"eval_completions/clipped_ratio": 0.0,
"eval_completions/max_length": 509.25,
"eval_completions/max_terminated_length": 509.25,
"eval_completions/mean_length": 219.36685180664062,
"eval_completions/mean_terminated_length": 219.36685180664062,
"eval_completions/min_length": 125.0,
"eval_completions/min_terminated_length": 125.0,
"eval_loss": 0.0,
"eval_num_tokens": 508303276.0,
"eval_reward": 0.792898640036583,
"eval_reward_std": 0.24335214123129845,
"eval_rewards/accuracy_reward": 0.421875,
"eval_rewards/batchaccgated_coverage_0": 0.03190650139003992,
"eval_rewards/batchaccgated_coverage_1": 0.03190650139003992,
"eval_rewards/batchaccgated_coverage_10": 0.031698619946837425,
"eval_rewards/batchaccgated_coverage_15": 0.030887323431670666,
"eval_rewards/batchaccgated_coverage_20": 0.027792250271886587,
"eval_rewards/batchaccgated_coverage_25": 0.025827288161963224,
"eval_rewards/batchaccgated_coverage_5": 0.03190650139003992,
"eval_rewards/brier_reward": 0.7553914785385132,
"eval_rewards/confidence_uniqueness_reward": 0.852294921875,
"eval_rewards/format_reward": 1.0,
"eval_rewards/frontier_entropy_batch_reward": -1.0,
"eval_runtime": 23.9378,
"eval_samples_per_second": 20.887,
"eval_signal/accuracy_reward/centered_abs_mean": 0.46875,
"eval_signal/accuracy_reward/group_std_mean": 0.49129971861839294,
"eval_signal/accuracy_reward/group_zero_std_frac": 0.0,
"eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9636169821023941,
"eval_signal/accuracy_reward/weight": 0.5,
"eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.234375,
"eval_signal/advantage_abs_mean": 0.9163325130939484,
"eval_signal/advantage_pre_scale_abs_mean": 0.2233411818742752,
"eval_signal/advantage_pre_scale_std": 0.24059945717453957,
"eval_signal/advantage_std": 0.9877118319272995,
"eval_signal/batchaccgated_coverage_0/centered_abs_mean": 0.05039947386831045,
"eval_signal/batchaccgated_coverage_0/group_std_mean": 0.06357055529952049,
"eval_signal/batchaccgated_coverage_0/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.0207137621473521,
"eval_signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.005039947456680238,
"eval_signal/batchaccgated_coverage_1/centered_abs_mean": 0.05039947386831045,
"eval_signal/batchaccgated_coverage_1/group_std_mean": 0.06357055529952049,
"eval_signal/batchaccgated_coverage_1/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.0207137621473521,
"eval_signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.005039947456680238,
"eval_signal/batchaccgated_coverage_10/centered_abs_mean": 0.05006645154207945,
"eval_signal/batchaccgated_coverage_10/group_std_mean": 0.06319636479020119,
"eval_signal/batchaccgated_coverage_10/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.02057558740489185,
"eval_signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.00500664534047246,
"eval_signal/batchaccgated_coverage_15/centered_abs_mean": 0.04867234081029892,
"eval_signal/batchaccgated_coverage_15/group_std_mean": 0.06159340962767601,
"eval_signal/batchaccgated_coverage_15/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.02000666386447847,
"eval_signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004867234267294407,
"eval_signal/batchaccgated_coverage_20/centered_abs_mean": 0.04169864021241665,
"eval_signal/batchaccgated_coverage_20/group_std_mean": 0.0532994968816638,
"eval_signal/batchaccgated_coverage_20/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.01712763891555369,
"eval_signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004169864114373922,
"eval_signal/batchaccgated_coverage_25/centered_abs_mean": 0.03730962937697768,
"eval_signal/batchaccgated_coverage_25/group_std_mean": 0.04770568013191223,
"eval_signal/batchaccgated_coverage_25/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.01531311310827732,
"eval_signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.003730962867848575,
"eval_signal/batchaccgated_coverage_5/centered_abs_mean": 0.05039947386831045,
"eval_signal/batchaccgated_coverage_5/group_std_mean": 0.06357055529952049,
"eval_signal/batchaccgated_coverage_5/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.0207137621473521,
"eval_signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.005039947456680238,
"eval_signal/brier_reward/centered_abs_mean": 0.27848223596811295,
"eval_signal/brier_reward/group_std_mean": 0.34658581018447876,
"eval_signal/brier_reward/group_zero_std_frac": 0.0,
"eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.11445964686572552,
"eval_signal/brier_reward/weight": 0.10000000149011612,
"eval_signal/brier_reward/weighted_centered_abs_mean": 0.02784822415560484,
"eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.08087158203125,
"eval_signal/confidence_uniqueness_reward/group_std_mean": 0.09585181809961796,
"eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.033175809774547815,
"eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.008087158203125,
"eval_signal/format_reward/centered_abs_mean": 0.0,
"eval_signal/format_reward/group_std_mean": 0.0,
"eval_signal/format_reward/group_zero_std_frac": 1.0,
"eval_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"eval_signal/format_reward/weight": 0.5,
"eval_signal/format_reward/weighted_centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"eval_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"eval_steps_per_second": 0.167,
"step": 150
},
{
"epoch": 0.48,
"step": 150,
"train_probe_completions/clipped_ratio": 0.002155172413793094,
"train_probe_completions/max_length": 415.5,
"train_probe_completions/max_terminated_length": 415.5,
"train_probe_completions/mean_length": 212.92254638671875,
"train_probe_completions/mean_terminated_length": 213.38474655151367,
"train_probe_completions/min_length": 91.75,
"train_probe_completions/min_terminated_length": 126.5,
"train_probe_loss": 0.0,
"train_probe_num_tokens": 508303276.0,
"train_probe_reward": 0.8844105452299118,
"train_probe_reward_std": 0.24425609409809113,
"train_probe_rewards/accuracy_reward": 0.587890625,
"train_probe_rewards/batchaccgated_coverage_0": 0.03745931992307305,
"train_probe_rewards/batchaccgated_coverage_1": 0.03745931992307305,
"train_probe_rewards/batchaccgated_coverage_10": 0.03442797111347318,
"train_probe_rewards/batchaccgated_coverage_15": 0.03793450025841594,
"train_probe_rewards/batchaccgated_coverage_20": 0.04565998539328575,
"train_probe_rewards/batchaccgated_coverage_25": 0.06084489822387695,
"train_probe_rewards/batchaccgated_coverage_5": 0.03745931992307305,
"train_probe_rewards/brier_reward": 0.8159727603197098,
"train_probe_rewards/confidence_uniqueness_reward": 0.8052466809749603,
"train_probe_rewards/format_reward": 0.998046875,
"train_probe_rewards/frontier_entropy_batch_reward": -0.998046875,
"train_probe_runtime": 29.191,
"train_probe_samples_per_second": 17.129,
"train_probe_signal/accuracy_reward/centered_abs_mean": 0.4725341796875,
"train_probe_signal/accuracy_reward/group_std_mean": 0.49347783625125885,
"train_probe_signal/accuracy_reward/group_zero_std_frac": 0.0,
"train_probe_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9684695601463318,
"train_probe_signal/accuracy_reward/weight": 0.5,
"train_probe_signal/accuracy_reward/weighted_centered_abs_mean": 0.23626708984375,
"train_probe_signal/advantage_abs_mean": 0.9070189148187637,
"train_probe_signal/advantage_pre_scale_abs_mean": 0.22153689712285995,
"train_probe_signal/advantage_pre_scale_std": 0.24152236431837082,
"train_probe_signal/advantage_std": 0.9877131432294846,
"train_probe_signal/batchaccgated_coverage_0/centered_abs_mean": 0.07761036418378353,
"train_probe_signal/batchaccgated_coverage_0/group_std_mean": 0.10870621353387833,
"train_probe_signal/batchaccgated_coverage_0/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.03194946702569723,
"train_probe_signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.007761036511510611,
"train_probe_signal/batchaccgated_coverage_1/centered_abs_mean": 0.07761036418378353,
"train_probe_signal/batchaccgated_coverage_1/group_std_mean": 0.10870621353387833,
"train_probe_signal/batchaccgated_coverage_1/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.03194946702569723,
"train_probe_signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.007761036511510611,
"train_probe_signal/batchaccgated_coverage_10/centered_abs_mean": 0.0687307883054018,
"train_probe_signal/batchaccgated_coverage_10/group_std_mean": 0.09697611257433891,
"train_probe_signal/batchaccgated_coverage_10/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.028283116407692432,
"train_probe_signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.006873079342767596,
"train_probe_signal/batchaccgated_coverage_15/centered_abs_mean": 0.059856778010725975,
"train_probe_signal/batchaccgated_coverage_15/group_std_mean": 0.08049065992236137,
"train_probe_signal/batchaccgated_coverage_15/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.024586224928498268,
"train_probe_signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.005985677824355662,
"train_probe_signal/batchaccgated_coverage_20/centered_abs_mean": 0.06132801156491041,
"train_probe_signal/batchaccgated_coverage_20/group_std_mean": 0.08038444258272648,
"train_probe_signal/batchaccgated_coverage_20/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.025221134535968304,
"train_probe_signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.006132801412604749,
"train_probe_signal/batchaccgated_coverage_25/centered_abs_mean": 0.0766581678763032,
"train_probe_signal/batchaccgated_coverage_25/group_std_mean": 0.09429071564227343,
"train_probe_signal/batchaccgated_coverage_25/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03159527666866779,
"train_probe_signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.007665816810913384,
"train_probe_signal/batchaccgated_coverage_5/centered_abs_mean": 0.07761036418378353,
"train_probe_signal/batchaccgated_coverage_5/group_std_mean": 0.10870621353387833,
"train_probe_signal/batchaccgated_coverage_5/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.03194946702569723,
"train_probe_signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.007761036511510611,
"train_probe_signal/brier_reward/centered_abs_mean": 0.22964580357074738,
"train_probe_signal/brier_reward/group_std_mean": 0.30229945480823517,
"train_probe_signal/brier_reward/group_zero_std_frac": 0.0,
"train_probe_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.09386581368744373,
"train_probe_signal/brier_reward/weight": 0.10000000149011612,
"train_probe_signal/brier_reward/weighted_centered_abs_mean": 0.022964580915868282,
"train_probe_signal/confidence_uniqueness_reward/centered_abs_mean": 0.13263628631830215,
"train_probe_signal/confidence_uniqueness_reward/group_std_mean": 0.15369696728885174,
"train_probe_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"train_probe_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.05450430233031511,
"train_probe_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"train_probe_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.013263628585264087,
"train_probe_signal/format_reward/centered_abs_mean": 0.0037841796875,
"train_probe_signal/format_reward/group_std_mean": 0.011048543266952038,
"train_probe_signal/format_reward/group_zero_std_frac": 0.9375,
"train_probe_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0074003045447170734,
"train_probe_signal/format_reward/weight": 0.5,
"train_probe_signal/format_reward/weighted_centered_abs_mean": 0.00189208984375,
"train_probe_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0037841796875,
"train_probe_signal/frontier_entropy_batch_reward/group_std_mean": 0.011048543266952038,
"train_probe_signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.9375,
"train_probe_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0014800610952079296,
"train_probe_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"train_probe_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0003784179862122983,
"train_probe_steps_per_second": 0.137
},
{
"calibration/aurc": 0.31140323081541144,
"calibration/batch_distribution_entropy": 0.7837272558968099,
"calibration/batch_entropy_100bins": 0.8117918102587524,
"calibration/batch_entropy_10bins": 0.7837272558968099,
"calibration/batch_entropy_50bins": 0.7969790531967555,
"calibration/batch_uniqueness": 0.8609405517578125,
"calibration/buffer_distribution_entropy": 0.8119208955554124,
"calibration/buffer_entropy_100bins": 0.857357276798006,
"calibration/buffer_entropy_10bins": 0.8119208955554124,
"calibration/buffer_entropy_50bins": 0.8363934716210212,
"calibration/confidence_entropy": 0.3045276418184897,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.2078125,
"calibration/coverage@15%": 0.277734375,
"calibration/coverage@20%": 0.347265625,
"calibration/coverage@25%": 0.388671875,
"calibration/coverage@30%": 0.519140625,
"calibration/coverage@5%": 0.18046875,
"calibration/distribution_entropy_10": 0.7837272558968099,
"calibration/distribution_entropy_100": 0.8117918102587524,
"calibration/ece": 0.2234017272051671,
"calibration/mean_confidence": 0.6814652306115916,
"calibration/unique_confidence_per_question": 0.803515625,
"calibration/unique_confidences": 411.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 677.2,
"completions/max_terminated_length": 677.2,
"completions/mean_length": 215.4796875,
"completions/mean_terminated_length": 215.52145385742188,
"completions/min_length": 64.8,
"completions/min_terminated_length": 106.4,
"epoch": 0.496,
"grad_norm": 0.009712236933410168,
"learning_rate": 1e-06,
"loss": -0.0034,
"num_tokens": 525817628.0,
"reward": 1.160148811340332,
"reward_std": 0.10452690571546555,
"rewards/accuracy_reward": 0.56650390625,
"rewards/batchaccgated_coverage_0": 0.3261030316352844,
"rewards/batchaccgated_coverage_1": 0.3261030316352844,
"rewards/batchaccgated_coverage_10": 0.38532768487930297,
"rewards/batchaccgated_coverage_15": 0.3903687298297882,
"rewards/batchaccgated_coverage_20": 0.3985093176364899,
"rewards/batchaccgated_coverage_25": 0.40261125564575195,
"rewards/batchaccgated_coverage_5": 0.36056922674179076,
"rewards/brier_reward": 0.787399423122406,
"rewards/confidence_uniqueness_reward": 0.8369665026664734,
"rewards/format_reward": 0.999609375,
"rewards/frontier_entropy_batch_reward": -0.4430363714694977,
"signal/accuracy_reward/centered_abs_mean": 0.085528564453125,
"signal/accuracy_reward/group_std_mean": 0.11582090556621552,
"signal/accuracy_reward/group_zero_std_frac": 0.659375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 3.049804139137268,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0427642822265625,
"signal/advantage_abs_mean": 0.7016374826431274,
"signal/advantage_pre_scale_abs_mean": 0.0749268189072609,
"signal/advantage_pre_scale_std": 0.14124129712581635,
"signal/advantage_std": 0.9776137828826904,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.03224932849407196,
"signal/batchaccgated_coverage_0/group_std_mean": 0.043821290135383606,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.2300462990999222,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.003224932961165905,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.03224932849407196,
"signal/batchaccgated_coverage_1/group_std_mean": 0.043821290135383606,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.2300462990999222,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.003224932961165905,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.04671005979180336,
"signal/batchaccgated_coverage_10/group_std_mean": 0.0665890820324421,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.3446627765893936,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004671006184071303,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.04907917231321335,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06996109634637833,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.359112361073494,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.00490791704505682,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.05512005165219307,
"signal/batchaccgated_coverage_20/group_std_mean": 0.07782004922628402,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.3986759543418884,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0055120051838457584,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.05970442593097687,
"signal/batchaccgated_coverage_25/group_std_mean": 0.08333772420883179,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.4399112641811371,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.00597044276073575,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03816870078444481,
"signal/batchaccgated_coverage_5/group_std_mean": 0.053555583208799364,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.28017411530017855,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0038168701343238355,
"signal/brier_reward/centered_abs_mean": 0.10170251876115799,
"signal/brier_reward/group_std_mean": 0.1334380328655243,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.7477420270442963,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010170252248644828,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.060911832749843596,
"signal/confidence_uniqueness_reward/group_std_mean": 0.07831140160560608,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.45565069019794463,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.006091183423995972,
"signal/format_reward/centered_abs_mean": 0.00074462890625,
"signal/format_reward/group_std_mean": 0.0018734002020210027,
"signal/format_reward/group_zero_std_frac": 0.990625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.03283944614231586,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000372314453125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2266017436981201,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2964478969573975,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.16875,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.6613954424858093,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02266017496585846,
"step": 155
},
{
"calibration/aurc": 0.2365859019867707,
"calibration/batch_distribution_entropy": 0.8312743318383106,
"calibration/batch_entropy_100bins": 0.8611416649727446,
"calibration/batch_entropy_10bins": 0.8312743318383106,
"calibration/batch_entropy_50bins": 0.8531621900205094,
"calibration/batch_uniqueness": 0.895489501953125,
"calibration/buffer_distribution_entropy": 0.8067217432995791,
"calibration/buffer_entropy_100bins": 0.8523879377409184,
"calibration/buffer_entropy_10bins": 0.8067217432995791,
"calibration/buffer_entropy_50bins": 0.830674928182875,
"calibration/confidence_entropy": 0.34841342275189724,
"calibration/coverage@0%": 0.0703125,
"calibration/coverage@1%": 0.0765625,
"calibration/coverage@10%": 0.35703125,
"calibration/coverage@15%": 0.426953125,
"calibration/coverage@20%": 0.487890625,
"calibration/coverage@25%": 0.54921875,
"calibration/coverage@30%": 0.65234375,
"calibration/coverage@5%": 0.24375,
"calibration/distribution_entropy_10": 0.8312743318383106,
"calibration/distribution_entropy_100": 0.8611416649727446,
"calibration/ece": 0.17004618005848665,
"calibration/mean_confidence": 0.6908863011772961,
"calibration/unique_confidence_per_question": 0.8625,
"calibration/unique_confidences": 441.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 564.0,
"completions/max_terminated_length": 564.0,
"completions/mean_length": 198.22275390625,
"completions/mean_terminated_length": 198.24056701660157,
"completions/min_length": 73.6,
"completions/min_terminated_length": 90.8,
"epoch": 0.512,
"grad_norm": 0.010849682614207268,
"learning_rate": 1e-06,
"loss": 0.0036,
"num_tokens": 542993093.0,
"reward": 1.1505912780761718,
"reward_std": 0.1027679294347763,
"rewards/accuracy_reward": 0.55166015625,
"rewards/batchaccgated_coverage_0": 0.31467311680316923,
"rewards/batchaccgated_coverage_1": 0.31467311680316923,
"rewards/batchaccgated_coverage_10": 0.377334451675415,
"rewards/batchaccgated_coverage_15": 0.3827280879020691,
"rewards/batchaccgated_coverage_20": 0.38887321949005127,
"rewards/batchaccgated_coverage_25": 0.39296114444732666,
"rewards/batchaccgated_coverage_5": 0.35811362862586976,
"rewards/brier_reward": 0.7968393206596375,
"rewards/confidence_uniqueness_reward": 0.85200115442276,
"rewards/format_reward": 0.99990234375,
"rewards/frontier_entropy_batch_reward": -0.430096960067749,
"signal/accuracy_reward/centered_abs_mean": 0.089471435546875,
"signal/accuracy_reward/group_std_mean": 0.12050010114908219,
"signal/accuracy_reward/group_zero_std_frac": 0.646875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 3.396079254150391,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0447357177734375,
"signal/advantage_abs_mean": 0.6996078133583069,
"signal/advantage_pre_scale_abs_mean": 0.07338147610425949,
"signal/advantage_pre_scale_std": 0.1404367983341217,
"signal/advantage_std": 0.9774126887321473,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.026165887713432312,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03600866869091988,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.19161012768745422,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0026165887946262957,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.026165887713432312,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03600866869091988,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.19161012768745422,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0026165887946262957,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03958204239606857,
"signal/batchaccgated_coverage_10/group_std_mean": 0.060008547455072406,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.30281769931316377,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.003958204435184598,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.042477143555879594,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06375285163521767,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.3223969578742981,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.00424771448597312,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04694569855928421,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06961265504360199,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.35201557278633117,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004694569949060678,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.051433315873146056,
"signal/batchaccgated_coverage_25/group_std_mean": 0.0748672679066658,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.3844266444444656,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.00514333164319396,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.033116813749074936,
"signal/batchaccgated_coverage_5/group_std_mean": 0.049293103814125064,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.24665699601173402,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003311681468039751,
"signal/brier_reward/centered_abs_mean": 0.10078437030315399,
"signal/brier_reward/group_std_mean": 0.1319650799036026,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.773439085483551,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010078437253832817,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.053031571209430695,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06679284498095513,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.41568334996700285,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.005303157120943069,
"signal/format_reward/centered_abs_mean": 0.000189208984375,
"signal/format_reward/group_std_mean": 0.0005524271633476019,
"signal/format_reward/group_zero_std_frac": 0.996875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.00448957309126854,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 9.46044921875e-05,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.21666131913661957,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.28842653036117555,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.14375,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.6231350898742676,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.021666132658720017,
"step": 160
},
{
"calibration/aurc": 0.20700501361615692,
"calibration/batch_distribution_entropy": 0.8030101181321964,
"calibration/batch_entropy_100bins": 0.8313590691552466,
"calibration/batch_entropy_10bins": 0.8030101181321964,
"calibration/batch_entropy_50bins": 0.8182401076774093,
"calibration/batch_uniqueness": 0.875588486474553,
"calibration/buffer_distribution_entropy": 0.8101174432225132,
"calibration/buffer_entropy_100bins": 0.8548773719117853,
"calibration/buffer_entropy_10bins": 0.8101174432225132,
"calibration/buffer_entropy_50bins": 0.8333923333049098,
"calibration/confidence_entropy": 0.3163789727210701,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.044140625,
"calibration/coverage@10%": 0.24140625,
"calibration/coverage@15%": 0.4305193014705882,
"calibration/coverage@20%": 0.5341038602941176,
"calibration/coverage@25%": 0.6115073529411765,
"calibration/coverage@30%": 0.7525735294117647,
"calibration/coverage@5%": 0.080078125,
"calibration/distribution_entropy_10": 0.8030101181321964,
"calibration/distribution_entropy_100": 0.8313590691552466,
"calibration/ece": 0.15397859387429086,
"calibration/mean_confidence": 0.7094141947903898,
"calibration/unique_confidence_per_question": 0.76171875,
"calibration/unique_confidences": 390.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 493.8,
"completions/max_terminated_length": 493.8,
"completions/mean_length": 178.82119140625,
"completions/mean_terminated_length": 178.85708312988282,
"completions/min_length": 71.6,
"completions/min_terminated_length": 89.0,
"epoch": 0.528,
"grad_norm": 0.013207187876105309,
"learning_rate": 1e-06,
"loss": -0.0029,
"num_tokens": 559853758.0,
"reward": 1.1485399007797241,
"reward_std": 0.09911583214998246,
"rewards/accuracy_reward": 0.54306640625,
"rewards/batchaccgated_coverage_0": 0.3212926506996155,
"rewards/batchaccgated_coverage_1": 0.3212926506996155,
"rewards/batchaccgated_coverage_10": 0.35536476969718933,
"rewards/batchaccgated_coverage_15": 0.3636810839176178,
"rewards/batchaccgated_coverage_20": 0.37284729480743406,
"rewards/batchaccgated_coverage_25": 0.37403364181518556,
"rewards/batchaccgated_coverage_5": 0.3402603268623352,
"rewards/brier_reward": 0.8224871277809143,
"rewards/confidence_uniqueness_reward": 0.8827354311943054,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.38295172452926635,
"signal/accuracy_reward/centered_abs_mean": 0.093243408203125,
"signal/accuracy_reward/group_std_mean": 0.12245183587074279,
"signal/accuracy_reward/group_zero_std_frac": 0.653125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.8007903575897215,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0466217041015625,
"signal/advantage_abs_mean": 0.7110128045082093,
"signal/advantage_pre_scale_abs_mean": 0.0741894707083702,
"signal/advantage_pre_scale_std": 0.1367669314146042,
"signal/advantage_std": 0.9787874460220337,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.028821926936507224,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03783415332436561,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.17249983549118042,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0028821928426623344,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.028821926936507224,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03783415332436561,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.17249983549118042,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0028821928426623344,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.037323671579360965,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05160753056406975,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.22395019829273224,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.003732367279008031,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.040981393307447433,
"signal/batchaccgated_coverage_15/group_std_mean": 0.056884602457284925,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.24415189325809478,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004098139191046357,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04693632051348686,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06479607820510865,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.277179542183876,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.00469363210722804,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04829187244176865,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06641278639435769,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.2863417088985443,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004829187132418156,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03332914300262928,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04491212144494057,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.19973039627075195,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003332914412021637,
"signal/brier_reward/centered_abs_mean": 0.09426200985908509,
"signal/brier_reward/group_std_mean": 0.12138466536998749,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.5623230993747711,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.00942620076239109,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.04248035848140717,
"signal/confidence_uniqueness_reward/group_std_mean": 0.05400928854942322,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.2537363857030869,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004248035978525877,
"signal/format_reward/centered_abs_mean": 0.0003662109375,
"signal/format_reward/group_std_mean": 0.000768545875325799,
"signal/format_reward/group_zero_std_frac": 0.996875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.009896960109472275,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.00018310546875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2157352477312088,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.28763124346733093,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.096875,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.2643781423568725,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02157352603971958,
"step": 165
},
{
"calibration/aurc": 0.19047471473856478,
"calibration/batch_distribution_entropy": 0.6684238529279624,
"calibration/batch_entropy_100bins": 0.734879728633403,
"calibration/batch_entropy_10bins": 0.6684238529279624,
"calibration/batch_entropy_50bins": 0.704239984676892,
"calibration/batch_uniqueness": 0.7910537033516108,
"calibration/buffer_distribution_entropy": 0.8072689781176706,
"calibration/buffer_entropy_100bins": 0.8534150978434447,
"calibration/buffer_entropy_10bins": 0.8072689781176706,
"calibration/buffer_entropy_50bins": 0.8313476964209606,
"calibration/confidence_entropy": 0.2581243895052179,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.29667701198630136,
"calibration/coverage@15%": 0.3752109833659491,
"calibration/coverage@20%": 0.5974949547455969,
"calibration/coverage@25%": 0.75415469055773,
"calibration/coverage@30%": 0.8162892918297455,
"calibration/coverage@5%": 0.07945205479452054,
"calibration/distribution_entropy_10": 0.6684238529279624,
"calibration/distribution_entropy_100": 0.734879728633403,
"calibration/ece": 0.1529910379179304,
"calibration/mean_confidence": 0.751871434850426,
"calibration/unique_confidence_per_question": 0.6046875,
"calibration/unique_confidences": 309.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 532.2,
"completions/max_terminated_length": 532.2,
"completions/mean_length": 179.5388671875,
"completions/mean_terminated_length": 179.556103515625,
"completions/min_length": 65.0,
"completions/min_terminated_length": 81.8,
"epoch": 0.544,
"grad_norm": 0.008139517158269882,
"learning_rate": 1e-06,
"loss": 0.0012,
"num_tokens": 576855820.0,
"reward": 1.1828856468200684,
"reward_std": 0.10724956393241883,
"rewards/accuracy_reward": 0.59541015625,
"rewards/batchaccgated_coverage_0": 0.34329968094825747,
"rewards/batchaccgated_coverage_1": 0.34329968094825747,
"rewards/batchaccgated_coverage_10": 0.40270864963531494,
"rewards/batchaccgated_coverage_15": 0.40859033465385436,
"rewards/batchaccgated_coverage_20": 0.4219518780708313,
"rewards/batchaccgated_coverage_25": 0.42657225728034975,
"rewards/batchaccgated_coverage_5": 0.3742012679576874,
"rewards/brier_reward": 0.7888158440589905,
"rewards/confidence_uniqueness_reward": 0.8312919735908508,
"rewards/format_reward": 0.99990234375,
"rewards/frontier_entropy_batch_reward": -0.4884375989437103,
"signal/accuracy_reward/centered_abs_mean": 0.099334716796875,
"signal/accuracy_reward/group_std_mean": 0.13290661275386811,
"signal/accuracy_reward/group_zero_std_frac": 0.615625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 3.87187123298645,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0496673583984375,
"signal/advantage_abs_mean": 0.7164512753486634,
"signal/advantage_pre_scale_abs_mean": 0.07729040384292603,
"signal/advantage_pre_scale_std": 0.14658820033073425,
"signal/advantage_std": 0.9779363751411438,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.025909633934497835,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03558695390820503,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.246875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.211699940264225,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0025909632910043,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.025909633934497835,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03558695390820503,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.246875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.211699940264225,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0025909632910043,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.04011059850454331,
"signal/batchaccgated_coverage_10/group_std_mean": 0.059988196939229965,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.246875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.34076022207736967,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004011060018092394,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.0428364422172308,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06378253772854806,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.246875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.3550789952278137,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.00428364435210824,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.052889463305473325,
"signal/batchaccgated_coverage_20/group_std_mean": 0.07746623381972313,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.246875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.42277740836143496,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.005288946395739913,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.05787998735904694,
"signal/batchaccgated_coverage_25/group_std_mean": 0.08347960337996482,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.246875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.4735015481710434,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.005787998717278242,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03133484907448292,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04528044611215591,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.246875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.2644213855266571,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003133485000580549,
"signal/brier_reward/centered_abs_mean": 0.09854650944471359,
"signal/brier_reward/group_std_mean": 0.12981645613908768,
"signal/brier_reward/group_zero_std_frac": 0.003125,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.7833450257778167,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009854651056230068,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.06094140559434891,
"signal/confidence_uniqueness_reward/group_std_mean": 0.07732685208320618,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.003125,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.5645584315061569,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0060941408388316635,
"signal/format_reward/centered_abs_mean": 0.000189208984375,
"signal/format_reward/group_std_mean": 0.0005524271633476019,
"signal/format_reward/group_zero_std_frac": 0.996875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.004211639612913131,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 9.46044921875e-05,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.20518638491630553,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.27104007005691527,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.20625,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.6775553345680236,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02051863893866539,
"step": 170
},
{
"calibration/aurc": 0.20614577137121431,
"calibration/batch_distribution_entropy": 0.7456864738570653,
"calibration/batch_entropy_100bins": 0.7981678869344162,
"calibration/batch_entropy_10bins": 0.7456864738570653,
"calibration/batch_entropy_50bins": 0.7783813950933718,
"calibration/batch_uniqueness": 0.852857352538468,
"calibration/buffer_distribution_entropy": 0.7972864133202512,
"calibration/buffer_entropy_100bins": 0.8473487130306807,
"calibration/buffer_entropy_10bins": 0.7972864133202512,
"calibration/buffer_entropy_50bins": 0.8238591960850655,
"calibration/confidence_entropy": 0.3061995079980194,
"calibration/coverage@0%": 0.0609375,
"calibration/coverage@1%": 0.108203125,
"calibration/coverage@10%": 0.376171875,
"calibration/coverage@15%": 0.4848183708414873,
"calibration/coverage@20%": 0.555570419520548,
"calibration/coverage@25%": 0.6282442514677105,
"calibration/coverage@30%": 0.6845141267123288,
"calibration/coverage@5%": 0.24609375,
"calibration/distribution_entropy_10": 0.7456864738570653,
"calibration/distribution_entropy_100": 0.7981678869344162,
"calibration/ece": 0.1778301072633955,
"calibration/mean_confidence": 0.7281486975880322,
"calibration/unique_confidence_per_question": 0.69453125,
"calibration/unique_confidences": 355.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 573.6,
"completions/max_terminated_length": 573.6,
"completions/mean_length": 182.7638671875,
"completions/mean_terminated_length": 182.79996337890626,
"completions/min_length": 74.4,
"completions/min_terminated_length": 93.6,
"epoch": 0.56,
"grad_norm": 0.009976216591894627,
"learning_rate": 1e-06,
"loss": 0.0013,
"num_tokens": 593548730.0,
"reward": 1.150138258934021,
"reward_std": 0.09659412056207657,
"rewards/accuracy_reward": 0.54091796875,
"rewards/batchaccgated_coverage_0": 0.32478708028793335,
"rewards/batchaccgated_coverage_1": 0.32478708028793335,
"rewards/batchaccgated_coverage_10": 0.3898737907409668,
"rewards/batchaccgated_coverage_15": 0.3991193413734436,
"rewards/batchaccgated_coverage_20": 0.40651315450668335,
"rewards/batchaccgated_coverage_25": 0.41008257269859316,
"rewards/batchaccgated_coverage_5": 0.3649914264678955,
"rewards/brier_reward": 0.7877224326133728,
"rewards/confidence_uniqueness_reward": 0.845093047618866,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.4552007257938385,
"signal/accuracy_reward/centered_abs_mean": 0.075677490234375,
"signal/accuracy_reward/group_std_mean": 0.10505554527044296,
"signal/accuracy_reward/group_zero_std_frac": 0.68125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 5.848579001426697,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0378387451171875,
"signal/advantage_abs_mean": 0.690128743648529,
"signal/advantage_pre_scale_abs_mean": 0.06619574204087257,
"signal/advantage_pre_scale_std": 0.13490325659513475,
"signal/advantage_std": 0.9737449169158936,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.02182765118777752,
"signal/batchaccgated_coverage_0/group_std_mean": 0.031006830930709838,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.3487917333841324,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0021827651653438807,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.02182765118777752,
"signal/batchaccgated_coverage_1/group_std_mean": 0.031006830930709838,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.3487917333841324,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0021827651653438807,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03637689277529717,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05749356374144554,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.49873943328857423,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0036376894451677797,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.040433880686759946,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06369718909263611,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.5785992264747619,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004043388273566961,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.046161182969808576,
"signal/batchaccgated_coverage_20/group_std_mean": 0.07140442132949829,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.699808943271637,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004616118362173438,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04985211044549942,
"signal/batchaccgated_coverage_25/group_std_mean": 0.07601254433393478,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.7742914021015167,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.0049852109514176846,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.02826472409069538,
"signal/batchaccgated_coverage_5/group_std_mean": 0.044160138443112375,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.4039997339248657,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0028264725115150213,
"signal/brier_reward/centered_abs_mean": 0.09340600073337554,
"signal/brier_reward/group_std_mean": 0.1257546365261078,
"signal/brier_reward/group_zero_std_frac": 0.009375,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 1.534515953063965,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009340600296854972,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.05425298064947128,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06879795342683792,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.009375,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.8427940040826798,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.005425298027694226,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.00872935652732849,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.20444384217262268,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.27165970802307127,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.175,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 3.325109100341797,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0204443845897913,
"step": 175
},
{
"calibration/aurc": 0.23387633983024364,
"calibration/batch_distribution_entropy": 0.7978808383897894,
"calibration/batch_entropy_100bins": 0.8121489744873811,
"calibration/batch_entropy_10bins": 0.7978808383897894,
"calibration/batch_entropy_50bins": 0.7938266464875896,
"calibration/batch_uniqueness": 0.8582330403748368,
"calibration/buffer_distribution_entropy": 0.7997895513119099,
"calibration/buffer_entropy_100bins": 0.848450671079388,
"calibration/buffer_entropy_10bins": 0.7997895513119099,
"calibration/buffer_entropy_50bins": 0.8247831374537595,
"calibration/confidence_entropy": 0.30822062250191484,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.25859375,
"calibration/coverage@15%": 0.4337366988747554,
"calibration/coverage@20%": 0.5075846991193738,
"calibration/coverage@25%": 0.5998004831213307,
"calibration/coverage@30%": 0.6873241805283757,
"calibration/coverage@5%": 0.046875,
"calibration/distribution_entropy_10": 0.7978808383897894,
"calibration/distribution_entropy_100": 0.8121489744873811,
"calibration/ece": 0.15204762162854019,
"calibration/mean_confidence": 0.6497209952064791,
"calibration/unique_confidence_per_question": 0.69375,
"calibration/unique_confidences": 355.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00029296875,
"completions/max_length": 475.8,
"completions/max_terminated_length": 475.8,
"completions/mean_length": 182.067578125,
"completions/mean_terminated_length": 182.12244262695313,
"completions/min_length": 54.2,
"completions/min_terminated_length": 88.2,
"epoch": 0.576,
"grad_norm": 0.009354103356599808,
"learning_rate": 1e-06,
"loss": -0.0018,
"num_tokens": 610599726.0,
"reward": 1.1488838195800781,
"reward_std": 0.09161649197340012,
"rewards/accuracy_reward": 0.541015625,
"rewards/batchaccgated_coverage_0": 0.3236666858196259,
"rewards/batchaccgated_coverage_1": 0.3236666858196259,
"rewards/batchaccgated_coverage_10": 0.3646405816078186,
"rewards/batchaccgated_coverage_15": 0.36668420433998106,
"rewards/batchaccgated_coverage_20": 0.37273460626602173,
"rewards/batchaccgated_coverage_25": 0.3748330414295197,
"rewards/batchaccgated_coverage_5": 0.3555976629257202,
"rewards/brier_reward": 0.8116917729377746,
"rewards/confidence_uniqueness_reward": 0.8797250270843506,
"rewards/format_reward": 0.99970703125,
"rewards/frontier_entropy_batch_reward": -0.3880158066749573,
"signal/accuracy_reward/centered_abs_mean": 0.07200927734375,
"signal/accuracy_reward/group_std_mean": 0.09560371786355973,
"signal/accuracy_reward/group_zero_std_frac": 0.721875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.2045305967330933,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.036004638671875,
"signal/advantage_abs_mean": 0.7145942211151123,
"signal/advantage_pre_scale_abs_mean": 0.06661182045936584,
"signal/advantage_pre_scale_std": 0.126612488925457,
"signal/advantage_std": 0.9786422610282898,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.032382354885339734,
"signal/batchaccgated_coverage_0/group_std_mean": 0.04306522309780121,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.340625,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.19750206768512726,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0032382356002926826,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.032382354885339734,
"signal/batchaccgated_coverage_1/group_std_mean": 0.04306522309780121,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.340625,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.19750206768512726,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0032382356002926826,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.04173615351319313,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05943812876939773,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.340625,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.2489446371793747,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004173615574836731,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.04267100691795349,
"signal/batchaccgated_coverage_15/group_std_mean": 0.060738421976566315,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.340625,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.2532739192247391,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004267100896686316,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04681762307882309,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06619012951850892,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.340625,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.27597851753234864,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004681762307882309,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04848678261041641,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06819486469030381,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.340625,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.2860581874847412,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004848678130656481,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03822807706892491,
"signal/batchaccgated_coverage_5/group_std_mean": 0.053860544413328174,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.340625,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.22971857488155364,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0038228077813982964,
"signal/brier_reward/centered_abs_mean": 0.08598587214946747,
"signal/brier_reward/group_std_mean": 0.11467064172029495,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.5178693056106567,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008598587289452553,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.041815857961773874,
"signal/confidence_uniqueness_reward/group_std_mean": 0.053715766221284864,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.2492111325263977,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004181585926562548,
"signal/format_reward/centered_abs_mean": 0.000567626953125,
"signal/format_reward/group_std_mean": 0.0016572814900428056,
"signal/format_reward/group_zero_std_frac": 0.990625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.016215188801288603,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0002838134765625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.22675967514514922,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.29980853796005247,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.121875,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.365171480178833,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02267596833407879,
"step": 180
},
{
"calibration/aurc": 0.2796563156622341,
"calibration/batch_distribution_entropy": 0.8491141704533529,
"calibration/batch_entropy_100bins": 0.8430700670558855,
"calibration/batch_entropy_10bins": 0.8491141704533529,
"calibration/batch_entropy_50bins": 0.834643771741917,
"calibration/batch_uniqueness": 0.876007675117196,
"calibration/buffer_distribution_entropy": 0.8039238237466652,
"calibration/buffer_entropy_100bins": 0.8502990618670747,
"calibration/buffer_entropy_10bins": 0.8039238237466652,
"calibration/buffer_entropy_50bins": 0.8266827186980535,
"calibration/confidence_entropy": 0.35184048288059844,
"calibration/coverage@0%": 0.050390625,
"calibration/coverage@1%": 0.116015625,
"calibration/coverage@10%": 0.31231194960861053,
"calibration/coverage@15%": 0.439457711594912,
"calibration/coverage@20%": 0.5062943370841487,
"calibration/coverage@25%": 0.5563379097358121,
"calibration/coverage@30%": 0.5997462084148728,
"calibration/coverage@5%": 0.2055849437377691,
"calibration/distribution_entropy_10": 0.8491141704533529,
"calibration/distribution_entropy_100": 0.8430700670558855,
"calibration/ece": 0.15801670659497874,
"calibration/mean_confidence": 0.6460033492536114,
"calibration/unique_confidence_per_question": 0.72890625,
"calibration/unique_confidences": 373.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 430.8,
"completions/max_terminated_length": 430.8,
"completions/mean_length": 182.79189453125,
"completions/mean_terminated_length": 182.81033325195312,
"completions/min_length": 70.8,
"completions/min_terminated_length": 90.2,
"epoch": 0.592,
"grad_norm": 0.008579729124903679,
"learning_rate": 1e-06,
"loss": -0.0006,
"num_tokens": 627639227.0,
"reward": 1.1398494243621826,
"reward_std": 0.09499209672212601,
"rewards/accuracy_reward": 0.5408203125,
"rewards/batchaccgated_coverage_0": 0.2984446585178375,
"rewards/batchaccgated_coverage_1": 0.2984446585178375,
"rewards/batchaccgated_coverage_10": 0.3482549965381622,
"rewards/batchaccgated_coverage_15": 0.35468419790267947,
"rewards/batchaccgated_coverage_20": 0.3597198724746704,
"rewards/batchaccgated_coverage_25": 0.36236393451690674,
"rewards/batchaccgated_coverage_5": 0.33214449882507324,
"rewards/brier_reward": 0.8060038924217224,
"rewards/confidence_uniqueness_reward": 0.8938992619514465,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.3585906982421875,
"signal/accuracy_reward/centered_abs_mean": 0.07305908203125,
"signal/accuracy_reward/group_std_mean": 0.10399322956800461,
"signal/accuracy_reward/group_zero_std_frac": 0.671875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 4.877646112442017,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.036529541015625,
"signal/advantage_abs_mean": 0.6944149017333985,
"signal/advantage_pre_scale_abs_mean": 0.06664436385035515,
"signal/advantage_pre_scale_std": 0.12623290866613388,
"signal/advantage_std": 0.9757420897483826,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.029301818832755087,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03870673254132271,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.37229459285736083,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0029301819391548633,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.029301818832755087,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03870673254132271,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.37229459285736083,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0029301819391548633,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03995690010488033,
"signal/batchaccgated_coverage_10/group_std_mean": 0.057380424439907075,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.5089192748069763,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.003995690122246742,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.04210545718669891,
"signal/batchaccgated_coverage_15/group_std_mean": 0.060422586649656294,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.5168418645858764,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.00421054563485086,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.044914111495018005,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06442278325557708,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.5296841710805893,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004491411196067929,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.0483915314078331,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06869830787181855,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.5833307504653931,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004839153215289116,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03490946888923645,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04941458180546761,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.328125,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.45007541179656985,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0034909469075500967,
"signal/brier_reward/centered_abs_mean": 0.08448402732610702,
"signal/brier_reward/group_std_mean": 0.11185397505760193,
"signal/brier_reward/group_zero_std_frac": 0.00625,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 1.063339602947235,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008448402769863606,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.03748636282980442,
"signal/confidence_uniqueness_reward/group_std_mean": 0.04786058068275452,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.00625,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.4530136674642563,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0037486365530639888,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.04034639894962311,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.22331035435199736,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.29462003111839297,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.128125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 2.7463122844696044,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.022331036254763605,
"step": 185
},
{
"calibration/aurc": 0.17985984264344515,
"calibration/batch_distribution_entropy": 0.7990657148900413,
"calibration/batch_entropy_100bins": 0.8362369636068788,
"calibration/batch_entropy_10bins": 0.7990657148900413,
"calibration/batch_entropy_50bins": 0.8228082036280041,
"calibration/batch_uniqueness": 0.8868804931640625,
"calibration/buffer_distribution_entropy": 0.8027568092840347,
"calibration/buffer_entropy_100bins": 0.848535746516295,
"calibration/buffer_entropy_10bins": 0.8027568092840347,
"calibration/buffer_entropy_50bins": 0.8245725445680222,
"calibration/confidence_entropy": 0.3312507705535863,
"calibration/coverage@0%": 0.101171875,
"calibration/coverage@1%": 0.126171875,
"calibration/coverage@10%": 0.4328125,
"calibration/coverage@15%": 0.5015625,
"calibration/coverage@20%": 0.60234375,
"calibration/coverage@25%": 0.68125,
"calibration/coverage@30%": 0.766015625,
"calibration/coverage@5%": 0.34140625,
"calibration/distribution_entropy_10": 0.7990657148900413,
"calibration/distribution_entropy_100": 0.8362369636068788,
"calibration/ece": 0.1431726052535786,
"calibration/mean_confidence": 0.679487407284143,
"calibration/unique_confidence_per_question": 0.75,
"calibration/unique_confidences": 384.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 511.4,
"completions/max_terminated_length": 511.4,
"completions/mean_length": 189.38515625,
"completions/mean_terminated_length": 189.4036071777344,
"completions/min_length": 80.2,
"completions/min_terminated_length": 101.8,
"epoch": 0.608,
"grad_norm": 0.008607161231338978,
"learning_rate": 1e-06,
"loss": 0.0008,
"num_tokens": 644578019.0,
"reward": 1.156525182723999,
"reward_std": 0.08738701641559601,
"rewards/accuracy_reward": 0.5337890625,
"rewards/batchaccgated_coverage_0": 0.33735418915748594,
"rewards/batchaccgated_coverage_1": 0.33735418915748594,
"rewards/batchaccgated_coverage_10": 0.38351939916610717,
"rewards/batchaccgated_coverage_15": 0.38584489226341245,
"rewards/batchaccgated_coverage_20": 0.390363872051239,
"rewards/batchaccgated_coverage_25": 0.39158823490142824,
"rewards/batchaccgated_coverage_5": 0.3703729212284088,
"rewards/brier_reward": 0.8181305527687073,
"rewards/confidence_uniqueness_reward": 0.8873992681503295,
"rewards/format_reward": 0.99990234375,
"rewards/frontier_entropy_batch_reward": -0.40513224601745607,
"signal/accuracy_reward/centered_abs_mean": 0.0743408203125,
"signal/accuracy_reward/group_std_mean": 0.10151887983083725,
"signal/accuracy_reward/group_zero_std_frac": 0.69375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.595080482959747,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.03717041015625,
"signal/advantage_abs_mean": 0.7050180077552796,
"signal/advantage_pre_scale_abs_mean": 0.06232886761426926,
"signal/advantage_pre_scale_std": 0.12167512625455856,
"signal/advantage_std": 0.9781908988952637,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.023159205541014673,
"signal/batchaccgated_coverage_0/group_std_mean": 0.030572206154465674,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.1607263579964638,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.002315920637920499,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.023159205541014673,
"signal/batchaccgated_coverage_1/group_std_mean": 0.030572206154465674,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.1607263579964638,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.002315920637920499,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03368106968700886,
"signal/batchaccgated_coverage_10/group_std_mean": 0.04895362704992294,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.2337779074907303,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0033681071829050778,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03485894165933132,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05068403333425522,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.24043053090572358,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.003485894203186035,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.03825772777199745,
"signal/batchaccgated_coverage_20/group_std_mean": 0.055334202200174334,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.26792340278625487,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.003825772972777486,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.03908900320529938,
"signal/batchaccgated_coverage_25/group_std_mean": 0.05642339363694191,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.27314221262931826,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.003908900311216712,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.029372266307473183,
"signal/batchaccgated_coverage_5/group_std_mean": 0.0415642224252224,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.2016366094350815,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0029372267425060274,
"signal/brier_reward/centered_abs_mean": 0.08187544345855713,
"signal/brier_reward/group_std_mean": 0.10775785893201828,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.5272607326507568,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008187544159591198,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.03658437728881836,
"signal/confidence_uniqueness_reward/group_std_mean": 0.046291432529687884,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.24734348356723784,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.003658437728881836,
"signal/format_reward/centered_abs_mean": 0.000189208984375,
"signal/format_reward/group_std_mean": 0.0005524271633476019,
"signal/format_reward/group_zero_std_frac": 0.996875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.012924566864967346,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 9.46044921875e-05,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.19728479981422425,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.27044069170951845,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.146875,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.2978215575218202,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.01972847953438759,
"step": 190
},
{
"calibration/aurc": 0.18774496656527173,
"calibration/batch_distribution_entropy": 0.7507878786547504,
"calibration/batch_entropy_100bins": 0.7784926337630923,
"calibration/batch_entropy_10bins": 0.7507878786547504,
"calibration/batch_entropy_50bins": 0.7612385221760973,
"calibration/batch_uniqueness": 0.8222991943359375,
"calibration/buffer_distribution_entropy": 0.8019993581938879,
"calibration/buffer_entropy_100bins": 0.8482786779214984,
"calibration/buffer_entropy_10bins": 0.8019993581938879,
"calibration/buffer_entropy_50bins": 0.8240946106804007,
"calibration/confidence_entropy": 0.2840769837529481,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.043359375,
"calibration/coverage@10%": 0.366796875,
"calibration/coverage@15%": 0.44296875,
"calibration/coverage@20%": 0.59765625,
"calibration/coverage@25%": 0.677734375,
"calibration/coverage@30%": 0.74609375,
"calibration/coverage@5%": 0.122265625,
"calibration/distribution_entropy_10": 0.7507878786547504,
"calibration/distribution_entropy_100": 0.7784926337630923,
"calibration/ece": 0.15453958109042432,
"calibration/mean_confidence": 0.7125681559779496,
"calibration/unique_confidence_per_question": 0.6578125,
"calibration/unique_confidences": 336.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 463.4,
"completions/max_terminated_length": 463.4,
"completions/mean_length": 195.0283203125,
"completions/mean_terminated_length": 195.0467498779297,
"completions/min_length": 81.2,
"completions/min_terminated_length": 102.8,
"epoch": 0.624,
"grad_norm": 0.01058657094836235,
"learning_rate": 1e-06,
"loss": 0.0005,
"num_tokens": 661919013.0,
"reward": 1.165335512161255,
"reward_std": 0.08911000788211823,
"rewards/accuracy_reward": 0.550390625,
"rewards/batchaccgated_coverage_0": 0.3535959303379059,
"rewards/batchaccgated_coverage_1": 0.3535959303379059,
"rewards/batchaccgated_coverage_10": 0.39958415031433103,
"rewards/batchaccgated_coverage_15": 0.4023648202419281,
"rewards/batchaccgated_coverage_20": 0.40959920883178713,
"rewards/batchaccgated_coverage_25": 0.4146316826343536,
"rewards/batchaccgated_coverage_5": 0.3810693323612213,
"rewards/brier_reward": 0.8045966982841491,
"rewards/confidence_uniqueness_reward": 0.8190199613571167,
"rewards/format_reward": 0.99990234375,
"rewards/frontier_entropy_batch_reward": -0.43616766929626466,
"signal/accuracy_reward/centered_abs_mean": 0.079931640625,
"signal/accuracy_reward/group_std_mean": 0.10740827172994613,
"signal/accuracy_reward/group_zero_std_frac": 0.684375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 4.038691067695618,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0399658203125,
"signal/advantage_abs_mean": 0.7172859311103821,
"signal/advantage_pre_scale_abs_mean": 0.0648710198700428,
"signal/advantage_pre_scale_std": 0.1260287046432495,
"signal/advantage_std": 0.976071584224701,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.020948295295238496,
"signal/batchaccgated_coverage_0/group_std_mean": 0.02843206450343132,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.2191994458436966,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.002094829687848687,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.020948295295238496,
"signal/batchaccgated_coverage_1/group_std_mean": 0.02843206450343132,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.2191994458436966,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.002094829687848687,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03147257007658481,
"signal/batchaccgated_coverage_10/group_std_mean": 0.04640885069966316,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.32031245827674865,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0031472570728510616,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03239230290055275,
"signal/batchaccgated_coverage_15/group_std_mean": 0.047355931252241135,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.32996806800365447,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0032392302993685005,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.036813342943787575,
"signal/batchaccgated_coverage_20/group_std_mean": 0.05351448729634285,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.36817744076251985,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0036813345737755298,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04129842668771744,
"signal/batchaccgated_coverage_25/group_std_mean": 0.05880364626646042,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.40780016779899597,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004129842761904002,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.025357672199606895,
"signal/batchaccgated_coverage_5/group_std_mean": 0.0360688079148531,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.2639189884066582,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0025357672944664954,
"signal/brier_reward/centered_abs_mean": 0.0875726118683815,
"signal/brier_reward/group_std_mean": 0.11331212967634201,
"signal/brier_reward/group_zero_std_frac": 0.003125,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.8669857084751129,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008757261373102666,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.0648462176322937,
"signal/confidence_uniqueness_reward/group_std_mean": 0.08304379284381866,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.003125,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.7208373367786407,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.006484622042626143,
"signal/format_reward/centered_abs_mean": 0.000189208984375,
"signal/format_reward/group_std_mean": 0.0005524271633476019,
"signal/format_reward/group_zero_std_frac": 0.996875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.019290830194950103,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 9.46044921875e-05,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.20197868049144746,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.265228682756424,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.190625,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 2.0048059701919554,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02019786797463894,
"step": 195
},
{
"calibration/aurc": 0.1915071915072601,
"calibration/batch_distribution_entropy": 0.7243413599012638,
"calibration/batch_entropy_100bins": 0.7939337287647554,
"calibration/batch_entropy_10bins": 0.7243413599012638,
"calibration/batch_entropy_50bins": 0.7732973050633243,
"calibration/batch_uniqueness": 0.8582585424399938,
"calibration/buffer_distribution_entropy": 0.8048248795288817,
"calibration/buffer_entropy_100bins": 0.8509430041840356,
"calibration/buffer_entropy_10bins": 0.8048248795288817,
"calibration/buffer_entropy_50bins": 0.8268919823792158,
"calibration/confidence_entropy": 0.2871044784052431,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0953125,
"calibration/coverage@10%": 0.34453125,
"calibration/coverage@15%": 0.490625,
"calibration/coverage@20%": 0.616919948630137,
"calibration/coverage@25%": 0.7001375978473581,
"calibration/coverage@30%": 0.7591433769569471,
"calibration/coverage@5%": 0.31484375,
"calibration/distribution_entropy_10": 0.7243413599012638,
"calibration/distribution_entropy_100": 0.7939337287647554,
"calibration/ece": 0.17423592856388043,
"calibration/mean_confidence": 0.7292497475878218,
"calibration/unique_confidence_per_question": 0.70546875,
"calibration/unique_confidences": 361.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.000390625,
"completions/max_length": 453.4,
"completions/max_terminated_length": 453.4,
"completions/mean_length": 185.98212890625,
"completions/mean_terminated_length": 186.0545654296875,
"completions/min_length": 63.2,
"completions/min_terminated_length": 102.4,
"epoch": 0.64,
"grad_norm": 0.009744562208652496,
"learning_rate": 1e-06,
"loss": 0.0012,
"num_tokens": 679166158.0,
"reward": 1.2284521341323853,
"reward_std": 0.07826440036296844,
"rewards/accuracy_reward": 0.5921875,
"rewards/batchaccgated_coverage_0": 0.42343791127204894,
"rewards/batchaccgated_coverage_1": 0.42343791127204894,
"rewards/batchaccgated_coverage_10": 0.45608340501785277,
"rewards/batchaccgated_coverage_15": 0.46087965965270994,
"rewards/batchaccgated_coverage_20": 0.464301598072052,
"rewards/batchaccgated_coverage_25": 0.4667300879955292,
"rewards/batchaccgated_coverage_5": 0.44429903030395507,
"rewards/brier_reward": 0.8170453429222106,
"rewards/confidence_uniqueness_reward": 0.8460218429565429,
"rewards/format_reward": 0.99951171875,
"rewards/frontier_entropy_batch_reward": -0.47621200084686277,
"signal/accuracy_reward/centered_abs_mean": 0.06229248046875,
"signal/accuracy_reward/group_std_mean": 0.08572121113538742,
"signal/accuracy_reward/group_zero_std_frac": 0.746875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 3.3950403213500975,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.031146240234375,
"signal/advantage_abs_mean": 0.7150399088859558,
"signal/advantage_pre_scale_abs_mean": 0.054778308421373365,
"signal/advantage_pre_scale_std": 0.11624106019735336,
"signal/advantage_std": 0.9743759751319885,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.025895629078149796,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03515827506780624,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.2591594636440277,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.002589562954381108,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.025895629078149796,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03515827506780624,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.2591594636440277,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.002589562954381108,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.032486167177557944,
"signal/batchaccgated_coverage_10/group_std_mean": 0.0475114606320858,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.3269293040037155,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0032486166339367626,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03490976020693779,
"signal/batchaccgated_coverage_15/group_std_mean": 0.051063787937164304,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.35622248649597166,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0034909760113805534,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.03719118870794773,
"signal/batchaccgated_coverage_20/group_std_mean": 0.054135237634181974,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.3915842741727829,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.003719118796288967,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.03981674648821354,
"signal/batchaccgated_coverage_25/group_std_mean": 0.05726717486977577,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.409972095489502,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.003981674695387483,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.02852901965379715,
"signal/batchaccgated_coverage_5/group_std_mean": 0.041091950982809065,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.28125,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.2905248820781708,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0028529020491987467,
"signal/brier_reward/centered_abs_mean": 0.07295950874686241,
"signal/brier_reward/group_std_mean": 0.09816256016492844,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.8064456760883332,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.007295950967818499,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.049154790490865706,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06347569972276687,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.5923048347234726,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004915479198098182,
"signal/format_reward/centered_abs_mean": 0.000933837890625,
"signal/format_reward/group_std_mean": 0.002425827318802476,
"signal/format_reward/group_zero_std_frac": 0.9875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.04495142698287964,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0004669189453125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.18465115129947662,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.24792596697807312,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.23125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.993400502204895,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.01846511512994766,
"step": 200
},
{
"epoch": 0.64,
"eval_completions/clipped_ratio": 0.0,
"eval_completions/max_length": 322.75,
"eval_completions/max_terminated_length": 322.75,
"eval_completions/mean_length": 180.15779876708984,
"eval_completions/mean_terminated_length": 180.15779876708984,
"eval_completions/min_length": 112.0,
"eval_completions/min_terminated_length": 112.0,
"eval_loss": 0.0,
"eval_num_tokens": 679166158.0,
"eval_reward": 0.7874932438135147,
"eval_reward_std": 0.24411927163600922,
"eval_rewards/accuracy_reward": 0.43359375,
"eval_rewards/batchaccgated_coverage_0": 0.026662104297429323,
"eval_rewards/batchaccgated_coverage_1": 0.026662104297429323,
"eval_rewards/batchaccgated_coverage_10": 0.026363940443843603,
"eval_rewards/batchaccgated_coverage_15": 0.026291401591151953,
"eval_rewards/batchaccgated_coverage_20": 0.024043998681008816,
"eval_rewards/batchaccgated_coverage_25": 0.02246633218601346,
"eval_rewards/batchaccgated_coverage_5": 0.026662104297429323,
"eval_rewards/brier_reward": 0.725321352481842,
"eval_rewards/confidence_uniqueness_reward": 0.802490234375,
"eval_rewards/format_reward": 1.0,
"eval_rewards/frontier_entropy_batch_reward": -1.0,
"eval_runtime": 17.8871,
"eval_samples_per_second": 27.953,
"eval_signal/accuracy_reward/centered_abs_mean": 0.468994140625,
"eval_signal/accuracy_reward/group_std_mean": 0.49164093285799026,
"eval_signal/accuracy_reward/group_zero_std_frac": 0.0,
"eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9612159579992294,
"eval_signal/accuracy_reward/weight": 0.5,
"eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.2344970703125,
"eval_signal/advantage_abs_mean": 0.9134364277124405,
"eval_signal/advantage_pre_scale_abs_mean": 0.22319187968969345,
"eval_signal/advantage_pre_scale_std": 0.24134470894932747,
"eval_signal/advantage_std": 0.9877133369445801,
"eval_signal/batchaccgated_coverage_0/centered_abs_mean": 0.046630797907710075,
"eval_signal/batchaccgated_coverage_0/group_std_mean": 0.06231147702783346,
"eval_signal/batchaccgated_coverage_0/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.019122778438031673,
"eval_signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0046630799770355225,
"eval_signal/batchaccgated_coverage_1/centered_abs_mean": 0.046630797907710075,
"eval_signal/batchaccgated_coverage_1/group_std_mean": 0.06231147702783346,
"eval_signal/batchaccgated_coverage_1/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.019122778438031673,
"eval_signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0046630799770355225,
"eval_signal/batchaccgated_coverage_10/centered_abs_mean": 0.04602299630641937,
"eval_signal/batchaccgated_coverage_10/group_std_mean": 0.06151102762669325,
"eval_signal/batchaccgated_coverage_10/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.01887426059693098,
"eval_signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004602299653925002,
"eval_signal/batchaccgated_coverage_15/centered_abs_mean": 0.045887960121035576,
"eval_signal/batchaccgated_coverage_15/group_std_mean": 0.06132367346435785,
"eval_signal/batchaccgated_coverage_15/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.018819204531610012,
"eval_signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0045887959422543645,
"eval_signal/batchaccgated_coverage_20/centered_abs_mean": 0.04031353443861008,
"eval_signal/batchaccgated_coverage_20/group_std_mean": 0.05412923824042082,
"eval_signal/batchaccgated_coverage_20/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.01652649836614728,
"eval_signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004031353630125523,
"eval_signal/batchaccgated_coverage_25/centered_abs_mean": 0.03522432781755924,
"eval_signal/batchaccgated_coverage_25/group_std_mean": 0.04734478984028101,
"eval_signal/batchaccgated_coverage_25/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.014441509963944554,
"eval_signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.0035224329330958426,
"eval_signal/batchaccgated_coverage_5/centered_abs_mean": 0.046630797907710075,
"eval_signal/batchaccgated_coverage_5/group_std_mean": 0.06231147702783346,
"eval_signal/batchaccgated_coverage_5/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.019122778438031673,
"eval_signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0046630799770355225,
"eval_signal/brier_reward/centered_abs_mean": 0.30063824355602264,
"eval_signal/brier_reward/group_std_mean": 0.36017828434705734,
"eval_signal/brier_reward/group_zero_std_frac": 0.0,
"eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.12322334945201874,
"eval_signal/brier_reward/weight": 0.10000000149011612,
"eval_signal/brier_reward/weighted_centered_abs_mean": 0.030063824728131294,
"eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.13763427734375,
"eval_signal/confidence_uniqueness_reward/group_std_mean": 0.15768801420927048,
"eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.05638975743204355,
"eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.013763428200036287,
"eval_signal/format_reward/centered_abs_mean": 0.0,
"eval_signal/format_reward/group_std_mean": 0.0,
"eval_signal/format_reward/group_zero_std_frac": 1.0,
"eval_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"eval_signal/format_reward/weight": 0.5,
"eval_signal/format_reward/weighted_centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"eval_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"eval_steps_per_second": 0.224,
"step": 200
},
{
"epoch": 0.64,
"step": 200,
"train_probe_completions/clipped_ratio": 0.0,
"train_probe_completions/max_length": 307.75,
"train_probe_completions/max_terminated_length": 307.75,
"train_probe_completions/mean_length": 179.1173210144043,
"train_probe_completions/mean_terminated_length": 179.1173210144043,
"train_probe_completions/min_length": 107.0,
"train_probe_completions/min_terminated_length": 107.0,
"train_probe_loss": 0.0,
"train_probe_num_tokens": 679166158.0,
"train_probe_reward": 0.8885173052549362,
"train_probe_reward_std": 0.24326353520154953,
"train_probe_rewards/accuracy_reward": 0.6015625,
"train_probe_rewards/batchaccgated_coverage_0": 0.03870750032365322,
"train_probe_rewards/batchaccgated_coverage_1": 0.03870750032365322,
"train_probe_rewards/batchaccgated_coverage_10": 0.03496906813234091,
"train_probe_rewards/batchaccgated_coverage_15": 0.03543397644534707,
"train_probe_rewards/batchaccgated_coverage_20": 0.0492563652805984,
"train_probe_rewards/batchaccgated_coverage_25": 0.07085349597036839,
"train_probe_rewards/batchaccgated_coverage_5": 0.03736709337681532,
"train_probe_rewards/brier_reward": 0.822309672832489,
"train_probe_rewards/confidence_uniqueness_reward": 0.749755859375,
"train_probe_rewards/format_reward": 1.0,
"train_probe_rewards/frontier_entropy_batch_reward": -1.0,
"train_probe_runtime": 17.2561,
"train_probe_samples_per_second": 28.975,
"train_probe_signal/accuracy_reward/centered_abs_mean": 0.467041015625,
"train_probe_signal/accuracy_reward/group_std_mean": 0.49060001224279404,
"train_probe_signal/accuracy_reward/group_zero_std_frac": 0.0,
"train_probe_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9623316824436188,
"train_probe_signal/accuracy_reward/weight": 0.5,
"train_probe_signal/accuracy_reward/weighted_centered_abs_mean": 0.2335205078125,
"train_probe_signal/advantage_abs_mean": 0.9091777503490448,
"train_probe_signal/advantage_pre_scale_abs_mean": 0.22102642059326172,
"train_probe_signal/advantage_pre_scale_std": 0.24054545909166336,
"train_probe_signal/advantage_std": 0.9877114146947861,
"train_probe_signal/batchaccgated_coverage_0/centered_abs_mean": 0.07323328778147697,
"train_probe_signal/batchaccgated_coverage_0/group_std_mean": 0.10610124096274376,
"train_probe_signal/batchaccgated_coverage_0/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.030201046261936426,
"train_probe_signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.007323328871279955,
"train_probe_signal/batchaccgated_coverage_1/centered_abs_mean": 0.07323328778147697,
"train_probe_signal/batchaccgated_coverage_1/group_std_mean": 0.10610124096274376,
"train_probe_signal/batchaccgated_coverage_1/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.030201046261936426,
"train_probe_signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.007323328871279955,
"train_probe_signal/batchaccgated_coverage_10/centered_abs_mean": 0.06401257123798132,
"train_probe_signal/batchaccgated_coverage_10/group_std_mean": 0.09422029741108418,
"train_probe_signal/batchaccgated_coverage_10/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.026412765961140394,
"train_probe_signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.006401257356628776,
"train_probe_signal/batchaccgated_coverage_15/centered_abs_mean": 0.04904160276055336,
"train_probe_signal/batchaccgated_coverage_15/group_std_mean": 0.07343923673033714,
"train_probe_signal/batchaccgated_coverage_15/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.020159508334472775,
"train_probe_signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.00490416050888598,
"train_probe_signal/batchaccgated_coverage_20/centered_abs_mean": 0.05487469490617514,
"train_probe_signal/batchaccgated_coverage_20/group_std_mean": 0.07427835371345282,
"train_probe_signal/batchaccgated_coverage_20/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.02238491363823414,
"train_probe_signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.005487469665240496,
"train_probe_signal/batchaccgated_coverage_25/centered_abs_mean": 0.077431321144104,
"train_probe_signal/batchaccgated_coverage_25/group_std_mean": 0.09515813272446394,
"train_probe_signal/batchaccgated_coverage_25/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.031520982505753636,
"train_probe_signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.007743132242467254,
"train_probe_signal/batchaccgated_coverage_5/centered_abs_mean": 0.06965978443622589,
"train_probe_signal/batchaccgated_coverage_5/group_std_mean": 0.10134058445692062,
"train_probe_signal/batchaccgated_coverage_5/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.028770192991942167,
"train_probe_signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.006965978536754847,
"train_probe_signal/brier_reward/centered_abs_mean": 0.22810199484229088,
"train_probe_signal/brier_reward/group_std_mean": 0.3001013919711113,
"train_probe_signal/brier_reward/group_zero_std_frac": 0.0,
"train_probe_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.09377521649003029,
"train_probe_signal/brier_reward/weight": 0.10000000149011612,
"train_probe_signal/brier_reward/weighted_centered_abs_mean": 0.022810200229287148,
"train_probe_signal/confidence_uniqueness_reward/centered_abs_mean": 0.178436279296875,
"train_probe_signal/confidence_uniqueness_reward/group_std_mean": 0.1961708664894104,
"train_probe_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"train_probe_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.07372372504323721,
"train_probe_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"train_probe_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.01784362830221653,
"train_probe_signal/format_reward/centered_abs_mean": 0.0,
"train_probe_signal/format_reward/group_std_mean": 0.0,
"train_probe_signal/format_reward/group_zero_std_frac": 1.0,
"train_probe_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/format_reward/weight": 0.5,
"train_probe_signal/format_reward/weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"train_probe_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"train_probe_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"train_probe_steps_per_second": 0.232
},
{
"calibration/aurc": 0.3504467524499325,
"calibration/batch_distribution_entropy": 0.7892451376786628,
"calibration/batch_entropy_100bins": 0.8321604451002689,
"calibration/batch_entropy_10bins": 0.7892451376786628,
"calibration/batch_entropy_50bins": 0.8181408654444937,
"calibration/batch_uniqueness": 0.8881011962890625,
"calibration/buffer_distribution_entropy": 0.8028593159038087,
"calibration/buffer_entropy_100bins": 0.8497270712668434,
"calibration/buffer_entropy_10bins": 0.8028593159038087,
"calibration/buffer_entropy_50bins": 0.8252664377789589,
"calibration/confidence_entropy": 0.32450192504505077,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.0,
"calibration/coverage@20%": 0.22734375,
"calibration/coverage@25%": 0.34140625,
"calibration/coverage@30%": 0.43984375,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.7892451376786628,
"calibration/distribution_entropy_100": 0.8321604451002689,
"calibration/ece": 0.27032662803267127,
"calibration/mean_confidence": 0.7064726392119038,
"calibration/unique_confidence_per_question": 0.727734375,
"calibration/unique_confidences": 372.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 358.4,
"completions/max_terminated_length": 358.4,
"completions/mean_length": 179.94482421875,
"completions/mean_terminated_length": 179.94482421875,
"completions/min_length": 96.4,
"completions/min_terminated_length": 96.4,
"epoch": 0.656,
"grad_norm": 0.008409109897911549,
"learning_rate": 1e-06,
"loss": -0.0025,
"num_tokens": 695865337.0,
"reward": 1.137472105026245,
"reward_std": 0.08662647008895874,
"rewards/accuracy_reward": 0.518359375,
"rewards/batchaccgated_coverage_0": 0.33675808310508726,
"rewards/batchaccgated_coverage_1": 0.33675808310508726,
"rewards/batchaccgated_coverage_10": 0.38846769332885744,
"rewards/batchaccgated_coverage_15": 0.39050665497779846,
"rewards/batchaccgated_coverage_20": 0.3920664608478546,
"rewards/batchaccgated_coverage_25": 0.39498435854911806,
"rewards/batchaccgated_coverage_5": 0.37108399271965026,
"rewards/brier_reward": 0.7640276193618775,
"rewards/confidence_uniqueness_reward": 0.845263671875,
"rewards/format_reward": 1.0,
"rewards/frontier_entropy_batch_reward": -0.43699292540550233,
"signal/accuracy_reward/centered_abs_mean": 0.06422119140625,
"signal/accuracy_reward/group_std_mean": 0.09017950594425202,
"signal/accuracy_reward/group_zero_std_frac": 0.71875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.9556093215942383,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.032110595703125,
"signal/advantage_abs_mean": 0.7084771752357483,
"signal/advantage_pre_scale_abs_mean": 0.06085373163223266,
"signal/advantage_pre_scale_std": 0.12351231276988983,
"signal/advantage_std": 0.9761632561683655,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.02272709459066391,
"signal/batchaccgated_coverage_0/group_std_mean": 0.030217368528246878,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.19770856201648712,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.002272709459066391,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.02272709459066391,
"signal/batchaccgated_coverage_1/group_std_mean": 0.030217368528246878,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.19770856201648712,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.002272709459066391,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.033816488459706306,
"signal/batchaccgated_coverage_10/group_std_mean": 0.050609685480594635,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.30604457557201387,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0033816488459706305,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.034772376716136935,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05194523110985756,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.3147363856434822,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.003477237839251757,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.0359696663916111,
"signal/batchaccgated_coverage_20/group_std_mean": 0.05345058962702751,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.32620810270309447,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0035969668067991735,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.03939540684223175,
"signal/batchaccgated_coverage_25/group_std_mean": 0.057603663951158526,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.3637551859021187,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.003939540823921561,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.028515692800283432,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04165886342525482,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.25829905569553374,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.002851569186896086,
"signal/brier_reward/centered_abs_mean": 0.09012337625026703,
"signal/brier_reward/group_std_mean": 0.11910613775253295,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.7986118495464325,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009012337774038315,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.052656030654907225,
"signal/confidence_uniqueness_reward/group_std_mean": 0.0656240276992321,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.4894114792346954,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.00526560302823782,
"signal/format_reward/centered_abs_mean": 0.0,
"signal/format_reward/group_std_mean": 0.0,
"signal/format_reward/group_zero_std_frac": 1.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.19708212912082673,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2594594031572342,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.2,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.7507773876190185,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.019708212837576865,
"step": 205
},
{
"calibration/aurc": 0.26466507923756166,
"calibration/batch_distribution_entropy": 0.7935069968910449,
"calibration/batch_entropy_100bins": 0.8262321369276175,
"calibration/batch_entropy_10bins": 0.7935069968910449,
"calibration/batch_entropy_50bins": 0.8117086557608861,
"calibration/batch_uniqueness": 0.886328125,
"calibration/buffer_distribution_entropy": 0.8034935668920781,
"calibration/buffer_entropy_100bins": 0.8515800643766755,
"calibration/buffer_entropy_10bins": 0.8034935668920781,
"calibration/buffer_entropy_50bins": 0.8272784548319112,
"calibration/confidence_entropy": 0.317329652698081,
"calibration/coverage@0%": 0.05546875,
"calibration/coverage@1%": 0.142578125,
"calibration/coverage@10%": 0.20546875,
"calibration/coverage@15%": 0.307421875,
"calibration/coverage@20%": 0.36328125,
"calibration/coverage@25%": 0.49140625,
"calibration/coverage@30%": 0.553125,
"calibration/coverage@5%": 0.172265625,
"calibration/distribution_entropy_10": 0.7935069968910449,
"calibration/distribution_entropy_100": 0.8262321369276175,
"calibration/ece": 0.18357197076710888,
"calibration/mean_confidence": 0.662241511154175,
"calibration/unique_confidence_per_question": 0.654296875,
"calibration/unique_confidences": 335.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 372.6,
"completions/max_terminated_length": 372.6,
"completions/mean_length": 177.77578125,
"completions/mean_terminated_length": 177.77578125,
"completions/min_length": 99.8,
"completions/min_terminated_length": 99.8,
"epoch": 0.672,
"grad_norm": 0.00860122125595808,
"learning_rate": 1e-06,
"loss": 0.0001,
"num_tokens": 712599201.0,
"reward": 1.1520925521850587,
"reward_std": 0.087205770611763,
"rewards/accuracy_reward": 0.5388671875,
"rewards/batchaccgated_coverage_0": 0.3397602796554565,
"rewards/batchaccgated_coverage_1": 0.3397602796554565,
"rewards/batchaccgated_coverage_10": 0.37700031995773314,
"rewards/batchaccgated_coverage_15": 0.38011021018028257,
"rewards/batchaccgated_coverage_20": 0.38551499843597414,
"rewards/batchaccgated_coverage_25": 0.3877726912498474,
"rewards/batchaccgated_coverage_5": 0.371553498506546,
"rewards/brier_reward": 0.8013934373855591,
"rewards/confidence_uniqueness_reward": 0.8661964416503907,
"rewards/format_reward": 1.0,
"rewards/frontier_entropy_batch_reward": -0.4224725246429443,
"signal/accuracy_reward/centered_abs_mean": 0.07679443359375,
"signal/accuracy_reward/group_std_mean": 0.10220114588737488,
"signal/accuracy_reward/group_zero_std_frac": 0.703125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.5142221689224242,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.038397216796875,
"signal/advantage_abs_mean": 0.7186357021331787,
"signal/advantage_pre_scale_abs_mean": 0.0634104035794735,
"signal/advantage_pre_scale_std": 0.12284806370735168,
"signal/advantage_std": 0.9781931519508362,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.029340996220707894,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03838393390178681,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.1895873323082924,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.002934099640697241,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.029340996220707894,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03838393390178681,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.1895873323082924,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.002934099640697241,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03704400286078453,
"signal/batchaccgated_coverage_10/group_std_mean": 0.051829658448696136,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.23600541055202484,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0037044002674520016,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03778132721781731,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05285651609301567,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.23959993422031403,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0037781327031552792,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04155103862285614,
"signal/batchaccgated_coverage_20/group_std_mean": 0.0579288050532341,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.2609895318746567,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004155103769153357,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04308544993400574,
"signal/batchaccgated_coverage_25/group_std_mean": 0.05997306555509567,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.27283562123775484,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004308545170351863,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03480848334729671,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04866644740104675,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.31875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.22416962683200836,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003480848530307412,
"signal/brier_reward/centered_abs_mean": 0.08313762992620469,
"signal/brier_reward/group_std_mean": 0.1090887352824211,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.5499769628047944,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008313762955367565,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.046487617492675784,
"signal/confidence_uniqueness_reward/group_std_mean": 0.056628217548131944,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.2939647018909454,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004648761730641127,
"signal/format_reward/centered_abs_mean": 0.0,
"signal/format_reward/group_std_mean": 0.0,
"signal/format_reward/group_zero_std_frac": 1.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2128327488899231,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2819286048412323,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.18125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.3950195789337159,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.021283274888992308,
"step": 210
},
{
"calibration/aurc": 0.2762679440302086,
"calibration/batch_distribution_entropy": 0.7844734099568491,
"calibration/batch_entropy_100bins": 0.8347366774595182,
"calibration/batch_entropy_10bins": 0.7844734099568491,
"calibration/batch_entropy_50bins": 0.8222736907104983,
"calibration/batch_uniqueness": 0.8911041259765625,
"calibration/buffer_distribution_entropy": 0.8027193710062152,
"calibration/buffer_entropy_100bins": 0.8519402357870502,
"calibration/buffer_entropy_10bins": 0.8027193710062152,
"calibration/buffer_entropy_50bins": 0.8276370512521553,
"calibration/confidence_entropy": 0.33750899099368614,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.217578125,
"calibration/coverage@15%": 0.274609375,
"calibration/coverage@20%": 0.432421875,
"calibration/coverage@25%": 0.492578125,
"calibration/coverage@30%": 0.570703125,
"calibration/coverage@5%": 0.091796875,
"calibration/distribution_entropy_10": 0.7844734099568491,
"calibration/distribution_entropy_100": 0.8347366774595182,
"calibration/ece": 0.20505037498649728,
"calibration/mean_confidence": 0.7259987949813158,
"calibration/unique_confidence_per_question": 0.666796875,
"calibration/unique_confidences": 341.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 360.2,
"completions/max_terminated_length": 360.2,
"completions/mean_length": 180.73212890625,
"completions/mean_terminated_length": 180.73212890625,
"completions/min_length": 102.8,
"completions/min_terminated_length": 102.8,
"epoch": 0.688,
"grad_norm": 0.011111167259514332,
"learning_rate": 1e-06,
"loss": 0.0006,
"num_tokens": 729403818.0,
"reward": 1.171406364440918,
"reward_std": 0.08351392447948455,
"rewards/accuracy_reward": 0.55849609375,
"rewards/batchaccgated_coverage_0": 0.34773980379104613,
"rewards/batchaccgated_coverage_1": 0.34773980379104613,
"rewards/batchaccgated_coverage_10": 0.386205267906189,
"rewards/batchaccgated_coverage_15": 0.3895800709724426,
"rewards/batchaccgated_coverage_20": 0.3971692085266113,
"rewards/batchaccgated_coverage_25": 0.39844772815704343,
"rewards/batchaccgated_coverage_5": 0.3720101475715637,
"rewards/brier_reward": 0.8021906614303589,
"rewards/confidence_uniqueness_reward": 0.8928070068359375,
"rewards/format_reward": 1.0,
"rewards/frontier_entropy_batch_reward": -0.4123064637184143,
"signal/accuracy_reward/centered_abs_mean": 0.072161865234375,
"signal/accuracy_reward/group_std_mean": 0.09350059926509857,
"signal/accuracy_reward/group_zero_std_frac": 0.734375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.0965453386306763,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0360809326171875,
"signal/advantage_abs_mean": 0.7240271091461181,
"signal/advantage_pre_scale_abs_mean": 0.061730723828077316,
"signal/advantage_pre_scale_std": 0.11806659549474716,
"signal/advantage_std": 0.9787090301513672,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.027986789867281914,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03643054887652397,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.1653071478009224,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.002798678958788514,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.027986789867281914,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03643054887652397,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.1653071478009224,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.002798678958788514,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03590458333492279,
"signal/batchaccgated_coverage_10/group_std_mean": 0.050425519794225694,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.21229459047317506,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.003590458445250988,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.037657585740089414,
"signal/batchaccgated_coverage_15/group_std_mean": 0.052764054387807846,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.22117720246315004,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0037657586857676507,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.042308029159903525,
"signal/batchaccgated_coverage_20/group_std_mean": 0.059184597432613374,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.2486230731010437,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004230803018435836,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.043913263082504275,
"signal/batchaccgated_coverage_25/group_std_mean": 0.061216804385185244,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.25772316157817843,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004391326243057847,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03171793632209301,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04344598278403282,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.1868104726076126,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.00317179374396801,
"signal/brier_reward/centered_abs_mean": 0.07320947349071502,
"signal/brier_reward/group_std_mean": 0.09601857960224151,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.4327149152755737,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.00732094757258892,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.039715695381164554,
"signal/confidence_uniqueness_reward/group_std_mean": 0.049058675020933154,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.23436517417430877,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.003971569798886776,
"signal/format_reward/centered_abs_mean": 0.0,
"signal/format_reward/group_std_mean": 0.0,
"signal/format_reward/group_zero_std_frac": 1.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.20287256836891174,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2679901719093323,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.153125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.2018382549285889,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.020287256687879562,
"step": 215
},
{
"calibration/aurc": 0.20851445085976578,
"calibration/batch_distribution_entropy": 0.6936041798504079,
"calibration/batch_entropy_100bins": 0.7508659452412398,
"calibration/batch_entropy_10bins": 0.6936041798504079,
"calibration/batch_entropy_50bins": 0.7253915132913564,
"calibration/batch_uniqueness": 0.8118944594255094,
"calibration/buffer_distribution_entropy": 0.7995465189599082,
"calibration/buffer_entropy_100bins": 0.8503417633687199,
"calibration/buffer_entropy_10bins": 0.7995465189599082,
"calibration/buffer_entropy_50bins": 0.8257138904654372,
"calibration/confidence_entropy": 0.27289649033917557,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.18612249266144815,
"calibration/coverage@15%": 0.3228794642857143,
"calibration/coverage@20%": 0.5482899645303327,
"calibration/coverage@25%": 0.6815114359099804,
"calibration/coverage@30%": 0.7666875611545988,
"calibration/coverage@5%": 0.07109375,
"calibration/distribution_entropy_10": 0.6936041798504079,
"calibration/distribution_entropy_100": 0.7508659452412398,
"calibration/ece": 0.20369775802245949,
"calibration/mean_confidence": 0.7617090104686435,
"calibration/unique_confidence_per_question": 0.5375,
"calibration/unique_confidences": 275.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 360.8,
"completions/max_terminated_length": 360.8,
"completions/mean_length": 177.25185546875,
"completions/mean_terminated_length": 177.26979370117186,
"completions/min_length": 80.6,
"completions/min_terminated_length": 101.4,
"epoch": 0.704,
"grad_norm": 0.009764812886714935,
"learning_rate": 1e-06,
"loss": 0.0001,
"num_tokens": 746085021.0,
"reward": 1.1820498704910278,
"reward_std": 0.08112008422613144,
"rewards/accuracy_reward": 0.558984375,
"rewards/batchaccgated_coverage_0": 0.38335577845573426,
"rewards/batchaccgated_coverage_1": 0.38335577845573426,
"rewards/batchaccgated_coverage_10": 0.43154327273368837,
"rewards/batchaccgated_coverage_15": 0.43594594597816466,
"rewards/batchaccgated_coverage_20": 0.4390959620475769,
"rewards/batchaccgated_coverage_25": 0.4410579323768616,
"rewards/batchaccgated_coverage_5": 0.41306591033935547,
"rewards/brier_reward": 0.7709994673728943,
"rewards/confidence_uniqueness_reward": 0.8191758155822754,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.49104241728782655,
"signal/accuracy_reward/centered_abs_mean": 0.06451416015625,
"signal/accuracy_reward/group_std_mean": 0.0869346410036087,
"signal/accuracy_reward/group_zero_std_frac": 0.74375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.127850317955017,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.032257080078125,
"signal/advantage_abs_mean": 0.7287756562232971,
"signal/advantage_pre_scale_abs_mean": 0.05786886215209961,
"signal/advantage_pre_scale_std": 0.12013832330703736,
"signal/advantage_std": 0.9782444119453431,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.01988573484122753,
"signal/batchaccgated_coverage_0/group_std_mean": 0.026858631521463394,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.126910237967968,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.001988573558628559,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.01988573484122753,
"signal/batchaccgated_coverage_1/group_std_mean": 0.026858631521463394,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.126910237967968,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.001988573558628559,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03085220456123352,
"signal/batchaccgated_coverage_10/group_std_mean": 0.046989626809954645,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.18945848792791367,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0030852204188704492,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03309219889342785,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05036740005016327,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.20494776964187622,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0033092197962105276,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.034732261672616005,
"signal/batchaccgated_coverage_20/group_std_mean": 0.052659335732460025,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.21629199385643005,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0034732263535261153,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.036953327059745786,
"signal/batchaccgated_coverage_25/group_std_mean": 0.05538719519972801,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.2283751040697098,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.003695332771167159,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.025232220813632012,
"signal/batchaccgated_coverage_5/group_std_mean": 0.037413445115089414,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.15674466490745545,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0025232221931219103,
"signal/brier_reward/centered_abs_mean": 0.07863166630268097,
"signal/brier_reward/group_std_mean": 0.10318497270345688,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.4951886415481567,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.007863166742026806,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.07369395196437836,
"signal/confidence_uniqueness_reward/group_std_mean": 0.08993445932865143,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.4484550267457962,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0073693949729204174,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.011796056851744651,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.17948388159275055,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.23878126144409179,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.20625,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.135391938686371,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.017948389053344727,
"step": 220
},
{
"calibration/aurc": 0.20306777029755887,
"calibration/batch_distribution_entropy": 0.7465173621964302,
"calibration/batch_entropy_100bins": 0.7936699400363958,
"calibration/batch_entropy_10bins": 0.7465173621964302,
"calibration/batch_entropy_50bins": 0.7725976228731111,
"calibration/batch_uniqueness": 0.860888671875,
"calibration/buffer_distribution_entropy": 0.7954736745841732,
"calibration/buffer_entropy_100bins": 0.8477467171584572,
"calibration/buffer_entropy_10bins": 0.7954736745841732,
"calibration/buffer_entropy_50bins": 0.8227282915303332,
"calibration/confidence_entropy": 0.28065822127000634,
"calibration/coverage@0%": 0.041796875,
"calibration/coverage@1%": 0.041796875,
"calibration/coverage@10%": 0.275,
"calibration/coverage@15%": 0.3328125,
"calibration/coverage@20%": 0.588671875,
"calibration/coverage@25%": 0.682421875,
"calibration/coverage@30%": 0.78515625,
"calibration/coverage@5%": 0.166796875,
"calibration/distribution_entropy_10": 0.7465173621964302,
"calibration/distribution_entropy_100": 0.7936699400363958,
"calibration/ece": 0.16865493260901077,
"calibration/mean_confidence": 0.7207870085417117,
"calibration/unique_confidence_per_question": 0.60390625,
"calibration/unique_confidences": 309.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 314.6,
"completions/max_terminated_length": 314.6,
"completions/mean_length": 171.00439453125,
"completions/mean_terminated_length": 171.00439453125,
"completions/min_length": 98.0,
"completions/min_terminated_length": 98.0,
"epoch": 0.72,
"grad_norm": 0.00839801412075758,
"learning_rate": 1e-06,
"loss": 0.0013,
"num_tokens": 762845962.0,
"reward": 1.1926531553268434,
"reward_std": 0.07956821620464324,
"rewards/accuracy_reward": 0.56708984375,
"rewards/batchaccgated_coverage_0": 0.3916622757911682,
"rewards/batchaccgated_coverage_1": 0.3916622757911682,
"rewards/batchaccgated_coverage_10": 0.4264684796333313,
"rewards/batchaccgated_coverage_15": 0.4276729762554169,
"rewards/batchaccgated_coverage_20": 0.43171672224998475,
"rewards/batchaccgated_coverage_25": 0.43402932286262513,
"rewards/batchaccgated_coverage_5": 0.41756036281585696,
"rewards/brier_reward": 0.797244393825531,
"rewards/confidence_uniqueness_reward": 0.8499969482421875,
"rewards/format_reward": 1.0,
"rewards/frontier_entropy_batch_reward": -0.4769319176673889,
"signal/accuracy_reward/centered_abs_mean": 0.066912841796875,
"signal/accuracy_reward/group_std_mean": 0.08809538185596466,
"signal/accuracy_reward/group_zero_std_frac": 0.74375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.8286802768707275,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0334564208984375,
"signal/advantage_abs_mean": 0.7215418577194214,
"signal/advantage_pre_scale_abs_mean": 0.05878787338733673,
"signal/advantage_pre_scale_std": 0.1158070683479309,
"signal/advantage_std": 0.9764286041259765,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.025292127579450607,
"signal/batchaccgated_coverage_0/group_std_mean": 0.033102256059646604,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.22594558894634248,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0025292127393186092,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.025292127579450607,
"signal/batchaccgated_coverage_1/group_std_mean": 0.033102256059646604,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.22594558894634248,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0025292127393186092,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.031799168512225154,
"signal/batchaccgated_coverage_10/group_std_mean": 0.04556064754724502,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.2759751111268997,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.003179916832596064,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03214199766516686,
"signal/batchaccgated_coverage_15/group_std_mean": 0.0460058443248272,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.27882693111896517,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0032141998875886203,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.03553076982498169,
"signal/batchaccgated_coverage_20/group_std_mean": 0.0504551112651825,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.30735174417495725,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0035530769266188143,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.03827302306890488,
"signal/batchaccgated_coverage_25/group_std_mean": 0.0537614144384861,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.32148835361003875,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.0038273023907095193,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.02942541241645813,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04172680899500847,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.2581797868013382,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.002942541381344199,
"signal/brier_reward/centered_abs_mean": 0.084199258685112,
"signal/brier_reward/group_std_mean": 0.10699962228536605,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.7347736418247223,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008419925905764103,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.05374593734741211,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06627482771873475,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.4973687916994095,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.00537459384649992,
"signal/format_reward/centered_abs_mean": 0.0,
"signal/format_reward/group_std_mean": 0.0,
"signal/format_reward/group_zero_std_frac": 1.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.19753661155700683,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.258336016535759,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.19375,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.6062188148498535,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.01975366100668907,
"step": 225
},
{
"calibration/aurc": 0.20053454591041692,
"calibration/batch_distribution_entropy": 0.7011184652918109,
"calibration/batch_entropy_100bins": 0.7590114244762448,
"calibration/batch_entropy_10bins": 0.7011184652918109,
"calibration/batch_entropy_50bins": 0.7340414442282165,
"calibration/batch_uniqueness": 0.8214813232421875,
"calibration/buffer_distribution_entropy": 0.7951408705112846,
"calibration/buffer_entropy_100bins": 0.8471364871387159,
"calibration/buffer_entropy_10bins": 0.7951408705112846,
"calibration/buffer_entropy_50bins": 0.8221355140639227,
"calibration/confidence_entropy": 0.2535950826043206,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.087109375,
"calibration/coverage@15%": 0.36640625,
"calibration/coverage@20%": 0.597265625,
"calibration/coverage@25%": 0.758984375,
"calibration/coverage@30%": 0.8625,
"calibration/coverage@5%": 0.06328125,
"calibration/distribution_entropy_10": 0.7011184652918109,
"calibration/distribution_entropy_100": 0.7590114244762448,
"calibration/ece": 0.17130612873248915,
"calibration/mean_confidence": 0.7379908394193192,
"calibration/unique_confidence_per_question": 0.56015625,
"calibration/unique_confidences": 286.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00029296875,
"completions/max_length": 354.6,
"completions/max_terminated_length": 354.6,
"completions/mean_length": 173.24423828125,
"completions/mean_terminated_length": 173.29540405273437,
"completions/min_length": 38.2,
"completions/min_terminated_length": 94.0,
"epoch": 0.736,
"grad_norm": 0.010154248215258121,
"learning_rate": 1e-06,
"loss": -0.0029,
"num_tokens": 779559567.0,
"reward": 1.2167637825012207,
"reward_std": 0.08232624530792236,
"rewards/accuracy_reward": 0.59521484375,
"rewards/batchaccgated_coverage_0": 0.3956462681293488,
"rewards/batchaccgated_coverage_1": 0.3956462681293488,
"rewards/batchaccgated_coverage_10": 0.4487145900726318,
"rewards/batchaccgated_coverage_15": 0.45863898992538454,
"rewards/batchaccgated_coverage_20": 0.4599409759044647,
"rewards/batchaccgated_coverage_25": 0.46182629466056824,
"rewards/batchaccgated_coverage_5": 0.4199801206588745,
"rewards/brier_reward": 0.8126120567321777,
"rewards/confidence_uniqueness_reward": 0.8278512001037598,
"rewards/format_reward": 0.99970703125,
"rewards/frontier_entropy_batch_reward": -0.48782858848571775,
"signal/accuracy_reward/centered_abs_mean": 0.059747314453125,
"signal/accuracy_reward/group_std_mean": 0.08410133123397827,
"signal/accuracy_reward/group_zero_std_frac": 0.740625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.8562824726104736,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0298736572265625,
"signal/advantage_abs_mean": 0.7195245027542114,
"signal/advantage_pre_scale_abs_mean": 0.05757363885641098,
"signal/advantage_pre_scale_std": 0.12290150225162506,
"signal/advantage_std": 0.9748034119606018,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.028381872922182083,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03737084940075874,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.2954946428537369,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.002838187455199659,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.028381872922182083,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03737084940075874,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.2954946428537369,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.002838187455199659,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.040372587740421295,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05897292420268059,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.4092389702796936,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004037258867174387,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.045624852925539014,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06707995533943176,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.4472480475902557,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004562485590577125,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04589327275753021,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06750083491206169,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.4552003383636475,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004589327517896891,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04709567427635193,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06904173120856286,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.4692850112915039,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.0047095675952732565,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03268024101853371,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04500489234924317,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.34388130605220796,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003268024232238531,
"signal/brier_reward/centered_abs_mean": 0.06939172521233558,
"signal/brier_reward/group_std_mean": 0.09392161816358566,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.6802172899246216,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.0069391726516187194,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.0637492373585701,
"signal/confidence_uniqueness_reward/group_std_mean": 0.07907815277576447,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.6140365660190582,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.006374923605471849,
"signal/format_reward/centered_abs_mean": 0.000567626953125,
"signal/format_reward/group_std_mean": 0.0016572814900428056,
"signal/format_reward/group_zero_std_frac": 0.990625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.02910785675048828,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0002838134765625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.1981547474861145,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.25699527859687804,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.23125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 2.052502727508545,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.01981547549366951,
"step": 230
},
{
"calibration/aurc": 0.20524579699700035,
"calibration/batch_distribution_entropy": 0.8014796385939965,
"calibration/batch_entropy_100bins": 0.8342311508969364,
"calibration/batch_entropy_10bins": 0.8014796385939965,
"calibration/batch_entropy_50bins": 0.8224307977397676,
"calibration/batch_uniqueness": 0.8920654296875,
"calibration/buffer_distribution_entropy": 0.7938525581992406,
"calibration/buffer_entropy_100bins": 0.846677364698885,
"calibration/buffer_entropy_10bins": 0.7938525581992406,
"calibration/buffer_entropy_50bins": 0.8217856370805958,
"calibration/confidence_entropy": 0.32126865333387855,
"calibration/coverage@0%": 0.058203125,
"calibration/coverage@1%": 0.091796875,
"calibration/coverage@10%": 0.391796875,
"calibration/coverage@15%": 0.45546875,
"calibration/coverage@20%": 0.5171875,
"calibration/coverage@25%": 0.63515625,
"calibration/coverage@30%": 0.73984375,
"calibration/coverage@5%": 0.24765625,
"calibration/distribution_entropy_10": 0.8014796385939965,
"calibration/distribution_entropy_100": 0.8342311508969364,
"calibration/ece": 0.18761018144748773,
"calibration/mean_confidence": 0.7038341147220594,
"calibration/unique_confidence_per_question": 0.6546875,
"calibration/unique_confidences": 335.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 412.8,
"completions/max_terminated_length": 412.8,
"completions/mean_length": 178.5638671875,
"completions/mean_terminated_length": 178.5638671875,
"completions/min_length": 100.6,
"completions/min_terminated_length": 100.6,
"epoch": 0.752,
"grad_norm": 0.00879575777798891,
"learning_rate": 1e-06,
"loss": 0.0001,
"num_tokens": 796615261.0,
"reward": 1.1970496654510498,
"reward_std": 0.08079780787229537,
"rewards/accuracy_reward": 0.57216796875,
"rewards/batchaccgated_coverage_0": 0.39651213884353637,
"rewards/batchaccgated_coverage_1": 0.39651213884353637,
"rewards/batchaccgated_coverage_10": 0.42684863805770873,
"rewards/batchaccgated_coverage_15": 0.4323661267757416,
"rewards/batchaccgated_coverage_20": 0.4370272934436798,
"rewards/batchaccgated_coverage_25": 0.44268366098403933,
"rewards/batchaccgated_coverage_5": 0.4129365861415863,
"rewards/brier_reward": 0.7881839513778687,
"rewards/confidence_uniqueness_reward": 0.8405265808105469,
"rewards/format_reward": 1.0,
"rewards/frontier_entropy_batch_reward": -0.46394086480140684,
"signal/accuracy_reward/centered_abs_mean": 0.061651611328125,
"signal/accuracy_reward/group_std_mean": 0.08349909484386445,
"signal/accuracy_reward/group_zero_std_frac": 0.75625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.1575896024703978,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0308258056640625,
"signal/advantage_abs_mean": 0.7235576510429382,
"signal/advantage_pre_scale_abs_mean": 0.05934170559048653,
"signal/advantage_pre_scale_std": 0.11974296569824219,
"signal/advantage_std": 0.9774991631507873,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.026659484207630157,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03521234057843685,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.18412656784057618,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0026659484719857574,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.026659484207630157,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03521234057843685,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.18412656784057618,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0026659484719857574,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.034586875885725024,
"signal/batchaccgated_coverage_10/group_std_mean": 0.04764077998697758,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.24197892248630523,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.003458687663078308,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03738788515329361,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05191471092402935,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.2614520609378815,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.003738788468763232,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04090946167707443,
"signal/batchaccgated_coverage_20/group_std_mean": 0.05663936585187912,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.2910120666027069,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0040909462608397,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04625504910945892,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06326544880867005,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.3386332869529724,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004625505115836859,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.029332654923200606,
"signal/batchaccgated_coverage_5/group_std_mean": 0.03918281942605972,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3125,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.20462170243263245,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0029332655016332866,
"signal/brier_reward/centered_abs_mean": 0.0875869944691658,
"signal/brier_reward/group_std_mean": 0.11215404570102691,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.6283963441848754,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.0087586997076869,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.052672076225280764,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06629086062312126,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.39270845651626585,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.005267207790166139,
"signal/format_reward/centered_abs_mean": 0.0,
"signal/format_reward/group_std_mean": 0.0,
"signal/format_reward/group_zero_std_frac": 1.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.19083771407604216,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.25120984613895414,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.228125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.3449601650238037,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.019083771109580993,
"step": 235
},
{
"calibration/aurc": 0.23219003720092682,
"calibration/batch_distribution_entropy": 0.7813207765190522,
"calibration/batch_entropy_100bins": 0.8293659224065705,
"calibration/batch_entropy_10bins": 0.7813207765190522,
"calibration/batch_entropy_50bins": 0.8152107363968792,
"calibration/batch_uniqueness": 0.8940948486328125,
"calibration/buffer_distribution_entropy": 0.7956969321825227,
"calibration/buffer_entropy_100bins": 0.8488144584443488,
"calibration/buffer_entropy_10bins": 0.7956969321825227,
"calibration/buffer_entropy_50bins": 0.8242839474704944,
"calibration/confidence_entropy": 0.3386048756803901,
"calibration/coverage@0%": 0.190625,
"calibration/coverage@1%": 0.209765625,
"calibration/coverage@10%": 0.29140625,
"calibration/coverage@15%": 0.337890625,
"calibration/coverage@20%": 0.37890625,
"calibration/coverage@25%": 0.51953125,
"calibration/coverage@30%": 0.673046875,
"calibration/coverage@5%": 0.24453125,
"calibration/distribution_entropy_10": 0.7813207765190522,
"calibration/distribution_entropy_100": 0.8293659224065705,
"calibration/ece": 0.18691760805622995,
"calibration/mean_confidence": 0.7342602120393716,
"calibration/unique_confidence_per_question": 0.66953125,
"calibration/unique_confidences": 342.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 384.4,
"completions/max_terminated_length": 384.4,
"completions/mean_length": 188.55771484375,
"completions/mean_terminated_length": 188.55771484375,
"completions/min_length": 105.0,
"completions/min_terminated_length": 105.0,
"epoch": 0.768,
"grad_norm": 0.011285772547125816,
"learning_rate": 1e-06,
"loss": -0.0004,
"num_tokens": 813478796.0,
"reward": 1.148850679397583,
"reward_std": 0.08340428471565246,
"rewards/accuracy_reward": 0.52216796875,
"rewards/batchaccgated_coverage_0": 0.34580066204071047,
"rewards/batchaccgated_coverage_1": 0.34580066204071047,
"rewards/batchaccgated_coverage_10": 0.3966225743293762,
"rewards/batchaccgated_coverage_15": 0.39749403595924376,
"rewards/batchaccgated_coverage_20": 0.400784569978714,
"rewards/batchaccgated_coverage_25": 0.40378488302230836,
"rewards/batchaccgated_coverage_5": 0.38525866270065307,
"rewards/brier_reward": 0.7776314496994019,
"rewards/confidence_uniqueness_reward": 0.8713752746582031,
"rewards/format_reward": 1.0,
"rewards/frontier_entropy_batch_reward": -0.44688597321510315,
"signal/accuracy_reward/centered_abs_mean": 0.061639404296875,
"signal/accuracy_reward/group_std_mean": 0.08581865727901458,
"signal/accuracy_reward/group_zero_std_frac": 0.7375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.7783201217651368,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0308197021484375,
"signal/advantage_abs_mean": 0.7383556246757508,
"signal/advantage_pre_scale_abs_mean": 0.059677806496620175,
"signal/advantage_pre_scale_std": 0.11951086074113845,
"signal/advantage_std": 0.9786966681480408,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.022053819522261618,
"signal/batchaccgated_coverage_0/group_std_mean": 0.028963499888777732,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.359375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.12853374183177949,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0022053821478039025,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.022053819522261618,
"signal/batchaccgated_coverage_1/group_std_mean": 0.028963499888777732,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.359375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.12853374183177949,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0022053821478039025,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03226460218429565,
"signal/batchaccgated_coverage_10/group_std_mean": 0.04766215085983276,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.359375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.18952254354953765,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0032264603767544033,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03272684551775455,
"signal/batchaccgated_coverage_15/group_std_mean": 0.048190195113420486,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.359375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.19304367154836655,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.003272684616968036,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.03500886857509613,
"signal/batchaccgated_coverage_20/group_std_mean": 0.05104769766330719,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.359375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.2055588647723198,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.003500887006521225,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.03882802575826645,
"signal/batchaccgated_coverage_25/group_std_mean": 0.05565792545676231,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.359375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.21996838450431824,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.0038828025572001934,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.028281287476420402,
"signal/batchaccgated_coverage_5/group_std_mean": 0.041326585412025454,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.359375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.16394689530134202,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.002828128868713975,
"signal/brier_reward/centered_abs_mean": 0.0867253378033638,
"signal/brier_reward/group_std_mean": 0.11273221224546433,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.4874584794044495,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008672533836215735,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.04427058696746826,
"signal/confidence_uniqueness_reward/group_std_mean": 0.05523076578974724,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.28006156384944914,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004427058901637792,
"signal/format_reward/centered_abs_mean": 0.0,
"signal/format_reward/group_std_mean": 0.0,
"signal/format_reward/group_zero_std_frac": 1.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2023990571498871,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2677549958229065,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.178125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.2018650770187378,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.020239906013011934,
"step": 240
},
{
"calibration/aurc": 0.24837061441683472,
"calibration/batch_distribution_entropy": 0.7497835725263522,
"calibration/batch_entropy_100bins": 0.7919657788318952,
"calibration/batch_entropy_10bins": 0.7497835725263522,
"calibration/batch_entropy_50bins": 0.7739817970728771,
"calibration/batch_uniqueness": 0.859771728515625,
"calibration/buffer_distribution_entropy": 0.7973821727007595,
"calibration/buffer_entropy_100bins": 0.8485112844239622,
"calibration/buffer_entropy_10bins": 0.7973821727007595,
"calibration/buffer_entropy_50bins": 0.824125943968242,
"calibration/confidence_entropy": 0.28750849797394307,
"calibration/coverage@0%": 0.0484375,
"calibration/coverage@1%": 0.121484375,
"calibration/coverage@10%": 0.263671875,
"calibration/coverage@15%": 0.405078125,
"calibration/coverage@20%": 0.446875,
"calibration/coverage@25%": 0.559765625,
"calibration/coverage@30%": 0.59921875,
"calibration/coverage@5%": 0.16953125,
"calibration/distribution_entropy_10": 0.7497835725263522,
"calibration/distribution_entropy_100": 0.7919657788318952,
"calibration/ece": 0.2058908066938859,
"calibration/mean_confidence": 0.704506798416857,
"calibration/unique_confidence_per_question": 0.61640625,
"calibration/unique_confidences": 315.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 438.6,
"completions/max_terminated_length": 438.6,
"completions/mean_length": 194.37587890625,
"completions/mean_terminated_length": 194.39525756835937,
"completions/min_length": 84.2,
"completions/min_terminated_length": 105.4,
"epoch": 0.784,
"grad_norm": 0.008082376793026924,
"learning_rate": 1e-06,
"loss": 0.0016,
"num_tokens": 830643573.0,
"reward": 1.1930540323257446,
"reward_std": 0.08683194518089295,
"rewards/accuracy_reward": 0.58701171875,
"rewards/batchaccgated_coverage_0": 0.35393508672714236,
"rewards/batchaccgated_coverage_1": 0.35393508672714236,
"rewards/batchaccgated_coverage_10": 0.4070340692996979,
"rewards/batchaccgated_coverage_15": 0.4126478016376495,
"rewards/batchaccgated_coverage_20": 0.41623960733413695,
"rewards/batchaccgated_coverage_25": 0.41737685799598695,
"rewards/batchaccgated_coverage_5": 0.388187313079834,
"rewards/brier_reward": 0.7964566469192504,
"rewards/confidence_uniqueness_reward": 0.8730056047439575,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.4223602056503296,
"signal/accuracy_reward/centered_abs_mean": 0.064984130859375,
"signal/accuracy_reward/group_std_mean": 0.09080888330936432,
"signal/accuracy_reward/group_zero_std_frac": 0.721875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 4.005117392539978,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0324920654296875,
"signal/advantage_abs_mean": 0.6930978655815124,
"signal/advantage_pre_scale_abs_mean": 0.0609614796936512,
"signal/advantage_pre_scale_std": 0.12167017608880996,
"signal/advantage_std": 0.9726897358894349,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.030781129375100136,
"signal/batchaccgated_coverage_0/group_std_mean": 0.04093157052993775,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.309375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.4209452599287033,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.003078112821094692,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.030781129375100136,
"signal/batchaccgated_coverage_1/group_std_mean": 0.04093157052993775,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.309375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.4209452599287033,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.003078112821094692,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.041489260643720625,
"signal/batchaccgated_coverage_10/group_std_mean": 0.06002146154642105,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.309375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.5561748802661896,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004148926353082061,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.04395114481449127,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06371254324913025,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.309375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.5860429346561432,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004395114723592997,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04646265655755997,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06706723496317864,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.309375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.6138898611068726,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004646265879273415,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04762714356184006,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06850429847836495,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.309375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.6261541426181794,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004762714356184005,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03718234300613403,
"signal/batchaccgated_coverage_5/group_std_mean": 0.052265846729278566,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.309375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.5079494595527649,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003718234319239855,
"signal/brier_reward/centered_abs_mean": 0.07728979289531708,
"signal/brier_reward/group_std_mean": 0.10346318483352661,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 1.0230805695056915,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.007728979270905256,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.0479205884039402,
"signal/confidence_uniqueness_reward/group_std_mean": 0.05968028232455254,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.6896383106708527,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004792058933526277,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.010998716205358505,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.20979842245578767,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.27545716762542727,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.140625,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 2.778763246536255,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02097984254360199,
"step": 245
},
{
"calibration/aurc": 0.16392369586853175,
"calibration/batch_distribution_entropy": 0.81710826291026,
"calibration/batch_entropy_100bins": 0.8410592656159228,
"calibration/batch_entropy_10bins": 0.81710826291026,
"calibration/batch_entropy_50bins": 0.8318963313664398,
"calibration/batch_uniqueness": 0.8930581232969029,
"calibration/buffer_distribution_entropy": 0.7956746697367063,
"calibration/buffer_entropy_100bins": 0.8473929249172428,
"calibration/buffer_entropy_10bins": 0.7956746697367063,
"calibration/buffer_entropy_50bins": 0.8229635492175229,
"calibration/confidence_entropy": 0.3247578873647905,
"calibration/coverage@0%": 0.15078125,
"calibration/coverage@1%": 0.224609375,
"calibration/coverage@10%": 0.4867715258192318,
"calibration/coverage@15%": 0.5462169915055063,
"calibration/coverage@20%": 0.6095686712184875,
"calibration/coverage@25%": 0.6697938310598212,
"calibration/coverage@30%": 0.7362598147135566,
"calibration/coverage@5%": 0.40500306372549016,
"calibration/distribution_entropy_10": 0.81710826291026,
"calibration/distribution_entropy_100": 0.8410592656159228,
"calibration/ece": 0.10639580803222559,
"calibration/mean_confidence": 0.6632614894288841,
"calibration/unique_confidence_per_question": 0.696875,
"calibration/unique_confidences": 356.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0005859375,
"completions/max_length": 484.8,
"completions/max_terminated_length": 484.8,
"completions/mean_length": 199.9166015625,
"completions/mean_terminated_length": 200.03225402832032,
"completions/min_length": 67.2,
"completions/min_terminated_length": 104.6,
"epoch": 0.8,
"grad_norm": 0.010074900463223457,
"learning_rate": 1e-06,
"loss": -0.0026,
"num_tokens": 847701279.0,
"reward": 1.2403445720672608,
"reward_std": 0.08357779979705811,
"rewards/accuracy_reward": 0.6162109375,
"rewards/batchaccgated_coverage_0": 0.40251862406730654,
"rewards/batchaccgated_coverage_1": 0.40251862406730654,
"rewards/batchaccgated_coverage_10": 0.45505309104919434,
"rewards/batchaccgated_coverage_15": 0.4630630314350128,
"rewards/batchaccgated_coverage_20": 0.4677554965019226,
"rewards/batchaccgated_coverage_25": 0.4690429031848907,
"rewards/batchaccgated_coverage_5": 0.44069035053253175,
"rewards/brier_reward": 0.8264482975006103,
"rewards/confidence_uniqueness_reward": 0.852568757534027,
"rewards/format_reward": 0.99931640625,
"rewards/frontier_entropy_batch_reward": -0.4538503348827362,
"signal/accuracy_reward/centered_abs_mean": 0.059716796875,
"signal/accuracy_reward/group_std_mean": 0.08749093711376191,
"signal/accuracy_reward/group_zero_std_frac": 0.71875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.8388143062591553,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0298583984375,
"signal/advantage_abs_mean": 0.7042630195617676,
"signal/advantage_pre_scale_abs_mean": 0.05688839554786682,
"signal/advantage_pre_scale_std": 0.11884100437164306,
"signal/advantage_std": 0.9781883120536804,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.03280194438993931,
"signal/batchaccgated_coverage_0/group_std_mean": 0.043618983775377276,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.265625,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.21338373720645903,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0032801944296807052,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.03280194438993931,
"signal/batchaccgated_coverage_1/group_std_mean": 0.043618983775377276,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.265625,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.21338373720645903,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0032801944296807052,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.043987677246332166,
"signal/batchaccgated_coverage_10/group_std_mean": 0.06215668097138405,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.265625,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.28277692198753357,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004398767650127411,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.047531963884830476,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06733132600784301,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.265625,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.3070178061723709,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.00475319642573595,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04935544058680534,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06994049400091171,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.265625,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.3196152806282043,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004935543984174728,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.05079603642225265,
"signal/batchaccgated_coverage_25/group_std_mean": 0.07172856032848358,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.265625,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.326677006483078,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.005079603474587202,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03874378278851509,
"signal/batchaccgated_coverage_5/group_std_mean": 0.054545413702726364,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.265625,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.2508065551519394,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0038743783254176377,
"signal/brier_reward/centered_abs_mean": 0.07452324032783508,
"signal/brier_reward/group_std_mean": 0.10029494762420654,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.4728826224803925,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.007452324219048023,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.0612126424908638,
"signal/confidence_uniqueness_reward/group_std_mean": 0.07577550858259201,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.39832741022109985,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.006121264398097992,
"signal/format_reward/centered_abs_mean": 0.001300048828125,
"signal/format_reward/group_std_mean": 0.0031943732872605326,
"signal/format_reward/group_zero_std_frac": 0.984375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.04635969698429108,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0006500244140625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.20528025925159454,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.27372945547103883,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.159375,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.317947220802307,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.020528025925159454,
"step": 250
},
{
"epoch": 0.8,
"eval_completions/clipped_ratio": 0.001953125,
"eval_completions/max_length": 377.25,
"eval_completions/max_terminated_length": 377.25,
"eval_completions/mean_length": 205.62041854858398,
"eval_completions/mean_terminated_length": 206.02280807495117,
"eval_completions/min_length": 94.5,
"eval_completions/min_terminated_length": 125.0,
"eval_loss": 0.0,
"eval_num_tokens": 847701279.0,
"eval_reward": 0.8084847331047058,
"eval_reward_std": 0.24739518389105797,
"eval_rewards/accuracy_reward": 0.45703125,
"eval_rewards/batchaccgated_coverage_0": 0.03194755269214511,
"eval_rewards/batchaccgated_coverage_1": 0.03194755269214511,
"eval_rewards/batchaccgated_coverage_10": 0.03176211332902312,
"eval_rewards/batchaccgated_coverage_15": 0.03161966102197766,
"eval_rewards/batchaccgated_coverage_20": 0.03184678731486201,
"eval_rewards/batchaccgated_coverage_25": 0.03216946450993419,
"eval_rewards/batchaccgated_coverage_5": 0.03194755269214511,
"eval_rewards/brier_reward": 0.7560671716928482,
"eval_rewards/confidence_uniqueness_reward": 0.8281958848237991,
"eval_rewards/format_reward": 0.998046875,
"eval_rewards/frontier_entropy_batch_reward": -0.998046875,
"eval_runtime": 31.0072,
"eval_samples_per_second": 16.125,
"eval_signal/accuracy_reward/centered_abs_mean": 0.480224609375,
"eval_signal/accuracy_reward/group_std_mean": 0.49776124954223633,
"eval_signal/accuracy_reward/group_zero_std_frac": 0.0,
"eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9730157405138016,
"eval_signal/accuracy_reward/weight": 0.5,
"eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.2401123046875,
"eval_signal/advantage_abs_mean": 0.922255203127861,
"eval_signal/advantage_pre_scale_abs_mean": 0.22821326553821564,
"eval_signal/advantage_pre_scale_std": 0.24479685351252556,
"eval_signal/advantage_std": 0.9877178817987442,
"eval_signal/batchaccgated_coverage_0/centered_abs_mean": 0.05604695528745651,
"eval_signal/batchaccgated_coverage_0/group_std_mean": 0.07248024642467499,
"eval_signal/batchaccgated_coverage_0/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.02271645748987794,
"eval_signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.005604695645160973,
"eval_signal/batchaccgated_coverage_1/centered_abs_mean": 0.05604695528745651,
"eval_signal/batchaccgated_coverage_1/group_std_mean": 0.07248024642467499,
"eval_signal/batchaccgated_coverage_1/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.02271645748987794,
"eval_signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.005604695645160973,
"eval_signal/batchaccgated_coverage_10/centered_abs_mean": 0.05559251829981804,
"eval_signal/batchaccgated_coverage_10/group_std_mean": 0.07190456613898277,
"eval_signal/batchaccgated_coverage_10/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.02253061020746827,
"eval_signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.005559251876547933,
"eval_signal/batchaccgated_coverage_15/centered_abs_mean": 0.05044085253030062,
"eval_signal/batchaccgated_coverage_15/group_std_mean": 0.06512190029025078,
"eval_signal/batchaccgated_coverage_15/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.020424326416105032,
"eval_signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.005044085322879255,
"eval_signal/batchaccgated_coverage_20/centered_abs_mean": 0.048408014699816704,
"eval_signal/batchaccgated_coverage_20/group_std_mean": 0.061856736429035664,
"eval_signal/batchaccgated_coverage_20/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.019615926314145327,
"eval_signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.00484080146998167,
"eval_signal/batchaccgated_coverage_25/centered_abs_mean": 0.04698999971151352,
"eval_signal/batchaccgated_coverage_25/group_std_mean": 0.05978013202548027,
"eval_signal/batchaccgated_coverage_25/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.01907806470990181,
"eval_signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004698999866377562,
"eval_signal/batchaccgated_coverage_5/centered_abs_mean": 0.05604695528745651,
"eval_signal/batchaccgated_coverage_5/group_std_mean": 0.07248024642467499,
"eval_signal/batchaccgated_coverage_5/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.02271645748987794,
"eval_signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.005604695645160973,
"eval_signal/brier_reward/centered_abs_mean": 0.27884741872549057,
"eval_signal/brier_reward/group_std_mean": 0.3426000103354454,
"eval_signal/brier_reward/group_zero_std_frac": 0.0,
"eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.11274541728198528,
"eval_signal/brier_reward/weight": 0.10000000149011612,
"eval_signal/brier_reward/weighted_centered_abs_mean": 0.027884742245078087,
"eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.10684499517083168,
"eval_signal/confidence_uniqueness_reward/group_std_mean": 0.12888846918940544,
"eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.043317739851772785,
"eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.010684499517083168,
"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": 0.007494867779314518,
"eval_signal/format_reward/weight": 0.5,
"eval_signal/format_reward/weighted_centered_abs_mean": 0.00189208984375,
"eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0037841796875,
"eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.011048543266952038,
"eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.9375,
"eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0014989734627306461,
"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": 0.129,
"step": 250
},
{
"epoch": 0.8,
"step": 250,
"train_probe_completions/clipped_ratio": 0.0,
"train_probe_completions/max_length": 330.75,
"train_probe_completions/max_terminated_length": 330.75,
"train_probe_completions/mean_length": 198.1436538696289,
"train_probe_completions/mean_terminated_length": 198.1436538696289,
"train_probe_completions/min_length": 118.0,
"train_probe_completions/min_terminated_length": 118.0,
"train_probe_loss": 0.0,
"train_probe_num_tokens": 847701279.0,
"train_probe_reward": 0.9041150063276291,
"train_probe_reward_std": 0.23714609071612358,
"train_probe_rewards/accuracy_reward": 0.623046875,
"train_probe_rewards/batchaccgated_coverage_0": 0.0409350348636508,
"train_probe_rewards/batchaccgated_coverage_1": 0.0409350348636508,
"train_probe_rewards/batchaccgated_coverage_10": 0.03647685516625643,
"train_probe_rewards/batchaccgated_coverage_15": 0.03967483341693878,
"train_probe_rewards/batchaccgated_coverage_20": 0.048978203907608986,
"train_probe_rewards/batchaccgated_coverage_25": 0.06688771769404411,
"train_probe_rewards/batchaccgated_coverage_5": 0.0409350348636508,
"train_probe_rewards/brier_reward": 0.8344815969467163,
"train_probe_rewards/confidence_uniqueness_reward": 0.776611328125,
"train_probe_rewards/format_reward": 1.0,
"train_probe_rewards/frontier_entropy_batch_reward": -1.0,
"train_probe_runtime": 18.4947,
"train_probe_samples_per_second": 27.035,
"train_probe_signal/accuracy_reward/centered_abs_mean": 0.4635009765625,
"train_probe_signal/accuracy_reward/group_std_mean": 0.48886142671108246,
"train_probe_signal/accuracy_reward/group_zero_std_frac": 0.0,
"train_probe_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.977079451084137,
"train_probe_signal/accuracy_reward/weight": 0.5,
"train_probe_signal/accuracy_reward/weighted_centered_abs_mean": 0.23175048828125,
"train_probe_signal/advantage_abs_mean": 0.8999495953321457,
"train_probe_signal/advantage_pre_scale_abs_mean": 0.2135889232158661,
"train_probe_signal/advantage_pre_scale_std": 0.2343948408961296,
"train_probe_signal/advantage_std": 0.9877015948295593,
"train_probe_signal/batchaccgated_coverage_0/centered_abs_mean": 0.08314423076808453,
"train_probe_signal/batchaccgated_coverage_0/group_std_mean": 0.11725048162043095,
"train_probe_signal/batchaccgated_coverage_0/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.03514544991776347,
"train_probe_signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.008314422448165715,
"train_probe_signal/batchaccgated_coverage_1/centered_abs_mean": 0.08314423076808453,
"train_probe_signal/batchaccgated_coverage_1/group_std_mean": 0.11725048162043095,
"train_probe_signal/batchaccgated_coverage_1/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.03514544991776347,
"train_probe_signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.008314422448165715,
"train_probe_signal/batchaccgated_coverage_10/centered_abs_mean": 0.06942837871611118,
"train_probe_signal/batchaccgated_coverage_10/group_std_mean": 0.10079094395041466,
"train_probe_signal/batchaccgated_coverage_10/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.029344776179641485,
"train_probe_signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.006942837848328054,
"train_probe_signal/batchaccgated_coverage_15/centered_abs_mean": 0.0568090733140707,
"train_probe_signal/batchaccgated_coverage_15/group_std_mean": 0.07850303314626217,
"train_probe_signal/batchaccgated_coverage_15/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.02396428305655718,
"train_probe_signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.005680907401256263,
"train_probe_signal/batchaccgated_coverage_20/centered_abs_mean": 0.05883408524096012,
"train_probe_signal/batchaccgated_coverage_20/group_std_mean": 0.0770224891602993,
"train_probe_signal/batchaccgated_coverage_20/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.024850953835994005,
"train_probe_signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.005883408593945205,
"train_probe_signal/batchaccgated_coverage_25/centered_abs_mean": 0.07662567123770714,
"train_probe_signal/batchaccgated_coverage_25/group_std_mean": 0.0928051145747304,
"train_probe_signal/batchaccgated_coverage_25/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03251631651073694,
"train_probe_signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.007662567310035229,
"train_probe_signal/batchaccgated_coverage_5/centered_abs_mean": 0.08314423076808453,
"train_probe_signal/batchaccgated_coverage_5/group_std_mean": 0.11725048162043095,
"train_probe_signal/batchaccgated_coverage_5/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.03514544991776347,
"train_probe_signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.008314422448165715,
"train_probe_signal/brier_reward/centered_abs_mean": 0.21461405977606773,
"train_probe_signal/brier_reward/group_std_mean": 0.28952909260988235,
"train_probe_signal/brier_reward/group_zero_std_frac": 0.0,
"train_probe_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.09034819900989532,
"train_probe_signal/brier_reward/weight": 0.10000000149011612,
"train_probe_signal/brier_reward/weighted_centered_abs_mean": 0.02146140579134226,
"train_probe_signal/confidence_uniqueness_reward/centered_abs_mean": 0.1571502685546875,
"train_probe_signal/confidence_uniqueness_reward/group_std_mean": 0.17540349811315536,
"train_probe_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"train_probe_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.06622758414596319,
"train_probe_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"train_probe_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.01571502722799778,
"train_probe_signal/format_reward/centered_abs_mean": 0.0,
"train_probe_signal/format_reward/group_std_mean": 0.0,
"train_probe_signal/format_reward/group_zero_std_frac": 1.0,
"train_probe_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/format_reward/weight": 0.5,
"train_probe_signal/format_reward/weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"train_probe_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"train_probe_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"train_probe_steps_per_second": 0.216
},
{
"calibration/aurc": 0.1983077339009832,
"calibration/batch_distribution_entropy": 0.6947641213794278,
"calibration/batch_entropy_100bins": 0.7560649489168251,
"calibration/batch_entropy_10bins": 0.6947641213794278,
"calibration/batch_entropy_50bins": 0.7335976365551428,
"calibration/batch_uniqueness": 0.8214788958216943,
"calibration/buffer_distribution_entropy": 0.7959595258919925,
"calibration/buffer_entropy_100bins": 0.8473740876640099,
"calibration/buffer_entropy_10bins": 0.7959595258919925,
"calibration/buffer_entropy_50bins": 0.8230819130828447,
"calibration/confidence_entropy": 0.26092542495024174,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.24101409313725491,
"calibration/coverage@15%": 0.29810968137254906,
"calibration/coverage@20%": 0.5473682598039216,
"calibration/coverage@25%": 0.6692984068627451,
"calibration/coverage@30%": 0.8107536764705883,
"calibration/coverage@5%": 0.19954044117647057,
"calibration/distribution_entropy_10": 0.6947641213794278,
"calibration/distribution_entropy_100": 0.7560649489168251,
"calibration/ece": 0.16748408280133767,
"calibration/mean_confidence": 0.739814860908098,
"calibration/unique_confidence_per_question": 0.584375,
"calibration/unique_confidences": 299.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0005859375,
"completions/max_length": 613.8,
"completions/max_terminated_length": 613.8,
"completions/mean_length": 201.80322265625,
"completions/mean_terminated_length": 201.92268981933594,
"completions/min_length": 42.8,
"completions/min_terminated_length": 107.6,
"epoch": 0.816,
"grad_norm": 0.008179843425750732,
"learning_rate": 1e-06,
"loss": -0.0022,
"num_tokens": 864866912.0,
"reward": 1.2196435451507568,
"reward_std": 0.08644702732563019,
"rewards/accuracy_reward": 0.59541015625,
"rewards/batchaccgated_coverage_0": 0.40730921030044553,
"rewards/batchaccgated_coverage_1": 0.40730921030044553,
"rewards/batchaccgated_coverage_10": 0.4478032231330872,
"rewards/batchaccgated_coverage_15": 0.4554340302944183,
"rewards/batchaccgated_coverage_20": 0.46207262873649596,
"rewards/batchaccgated_coverage_25": 0.463405179977417,
"rewards/batchaccgated_coverage_5": 0.43226310014724734,
"rewards/brier_reward": 0.7984767556190491,
"rewards/confidence_uniqueness_reward": 0.8349117875099182,
"rewards/format_reward": 0.9994140625,
"rewards/frontier_entropy_batch_reward": -0.4866710603237152,
"signal/accuracy_reward/centered_abs_mean": 0.069195556640625,
"signal/accuracy_reward/group_std_mean": 0.09374037981033326,
"signal/accuracy_reward/group_zero_std_frac": 0.725,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.374647521972656,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0345977783203125,
"signal/advantage_abs_mean": 0.7181798458099365,
"signal/advantage_pre_scale_abs_mean": 0.06134960576891899,
"signal/advantage_pre_scale_std": 0.12555146366357803,
"signal/advantage_std": 0.9777508735656738,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.027879495173692703,
"signal/batchaccgated_coverage_0/group_std_mean": 0.037643495947122574,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.19717624485492707,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0027879495173692703,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.027879495173692703,
"signal/batchaccgated_coverage_1/group_std_mean": 0.037643495947122574,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.19717624485492707,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0027879495173692703,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03705677837133407,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05376128479838371,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.2587757557630539,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0037056779488921165,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.04139879867434502,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06024034470319748,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.28523273169994356,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.00413988009095192,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.045519938319921495,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06580962762236595,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.3159133195877075,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004551994102075696,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04676199778914451,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06728066727519036,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.32522574365139006,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004676200030371547,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03178059011697769,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04519719034433365,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.22377659678459166,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0031780591700226067,
"signal/brier_reward/centered_abs_mean": 0.08286205679178238,
"signal/brier_reward/group_std_mean": 0.10989545434713363,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.5640357732772827,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008286205679178238,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.0629216730594635,
"signal/confidence_uniqueness_reward/group_std_mean": 0.07908432558178902,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.4394727200269699,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.006292167212814092,
"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": 0.0352968692779541,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0005615234375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2005961626768112,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.26307234168052673,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.1875,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.3928106069564818,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02005961649119854,
"step": 255
},
{
"calibration/aurc": 0.24010140134329355,
"calibration/batch_distribution_entropy": 0.743408482474009,
"calibration/batch_entropy_100bins": 0.8047442302288262,
"calibration/batch_entropy_10bins": 0.743408482474009,
"calibration/batch_entropy_50bins": 0.7841642055635922,
"calibration/batch_uniqueness": 0.8775421142578125,
"calibration/buffer_distribution_entropy": 0.7925249730038597,
"calibration/buffer_entropy_100bins": 0.8447734138203977,
"calibration/buffer_entropy_10bins": 0.7925249730038597,
"calibration/buffer_entropy_50bins": 0.8200888124740894,
"calibration/confidence_entropy": 0.30132111282457175,
"calibration/coverage@0%": 0.11015625,
"calibration/coverage@1%": 0.1859375,
"calibration/coverage@10%": 0.25078125,
"calibration/coverage@15%": 0.304296875,
"calibration/coverage@20%": 0.451953125,
"calibration/coverage@25%": 0.584765625,
"calibration/coverage@30%": 0.669921875,
"calibration/coverage@5%": 0.204296875,
"calibration/distribution_entropy_10": 0.743408482474009,
"calibration/distribution_entropy_100": 0.8047442302288262,
"calibration/ece": 0.21133256107536047,
"calibration/mean_confidence": 0.7262744319773995,
"calibration/unique_confidence_per_question": 0.6421875,
"calibration/unique_confidences": 328.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 704.8,
"completions/max_terminated_length": 704.8,
"completions/mean_length": 203.5560546875,
"completions/mean_terminated_length": 203.57622985839845,
"completions/min_length": 86.8,
"completions/min_terminated_length": 107.0,
"epoch": 0.832,
"grad_norm": 0.007974312640726566,
"learning_rate": 1e-06,
"loss": -0.0016,
"num_tokens": 881959678.0,
"reward": 1.1928089618682862,
"reward_std": 0.08918431848287582,
"rewards/accuracy_reward": 0.5689453125,
"rewards/batchaccgated_coverage_0": 0.3751842677593231,
"rewards/batchaccgated_coverage_1": 0.3751842677593231,
"rewards/batchaccgated_coverage_10": 0.42907959818840025,
"rewards/batchaccgated_coverage_15": 0.43304158449172975,
"rewards/batchaccgated_coverage_20": 0.4355816125869751,
"rewards/batchaccgated_coverage_25": 0.4377098441123962,
"rewards/batchaccgated_coverage_5": 0.40600188374519347,
"rewards/brier_reward": 0.8024233460426331,
"rewards/confidence_uniqueness_reward": 0.8435166835784912,
"rewards/format_reward": 0.99990234375,
"rewards/frontier_entropy_batch_reward": -0.4538717806339264,
"signal/accuracy_reward/centered_abs_mean": 0.07647705078125,
"signal/accuracy_reward/group_std_mean": 0.10265623033046722,
"signal/accuracy_reward/group_zero_std_frac": 0.69375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.1891923427581785,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.038238525390625,
"signal/advantage_abs_mean": 0.7312221884727478,
"signal/advantage_pre_scale_abs_mean": 0.06450615003705025,
"signal/advantage_pre_scale_std": 0.12873434722423555,
"signal/advantage_std": 0.9788714170455932,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.024187516793608664,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03238842785358429,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.13592786639928817,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0024187516886740924,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.024187516793608664,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03238842785358429,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.13592786639928817,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0024187516886740924,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.035252517089247704,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05173556432127953,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.19932306408882142,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0035252516623586415,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03709460571408272,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05454001948237419,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.21038373410701752,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0037094606552273033,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.03888828344643116,
"signal/batchaccgated_coverage_20/group_std_mean": 0.05701977014541626,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.2217923641204834,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0038888282142579556,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04146089106798172,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06009025797247887,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.23793689608573915,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004146089218556881,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.029069171473383903,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04164224863052368,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.2875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.1646407127380371,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0029069172218441964,
"signal/brier_reward/centered_abs_mean": 0.09380194991827011,
"signal/brier_reward/group_std_mean": 0.12048914581537247,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.537421065568924,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009380194917321205,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.05585924088954926,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06987837180495263,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.3221887081861496,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.005585924163460732,
"signal/format_reward/centered_abs_mean": 0.000189208984375,
"signal/format_reward/group_std_mean": 0.0005524271633476019,
"signal/format_reward/group_zero_std_frac": 0.996875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.006292828917503357,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 9.46044921875e-05,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.18837104141712188,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.24778571128845214,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.209375,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.0876347303390503,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.018837105110287668,
"step": 260
},
{
"calibration/aurc": 0.2725763855039081,
"calibration/batch_distribution_entropy": 0.70133045972446,
"calibration/batch_entropy_100bins": 0.763734786097396,
"calibration/batch_entropy_10bins": 0.70133045972446,
"calibration/batch_entropy_50bins": 0.7410124854466985,
"calibration/batch_uniqueness": 0.83316650390625,
"calibration/buffer_distribution_entropy": 0.7892171682897671,
"calibration/buffer_entropy_100bins": 0.842852457305073,
"calibration/buffer_entropy_10bins": 0.7892171682897671,
"calibration/buffer_entropy_50bins": 0.8178552953238981,
"calibration/confidence_entropy": 0.26855049016011745,
"calibration/coverage@0%": 0.083203125,
"calibration/coverage@1%": 0.094140625,
"calibration/coverage@10%": 0.223828125,
"calibration/coverage@15%": 0.267578125,
"calibration/coverage@20%": 0.455859375,
"calibration/coverage@25%": 0.494921875,
"calibration/coverage@30%": 0.56171875,
"calibration/coverage@5%": 0.1296875,
"calibration/distribution_entropy_10": 0.70133045972446,
"calibration/distribution_entropy_100": 0.763734786097396,
"calibration/ece": 0.21207872787594884,
"calibration/mean_confidence": 0.741867364970565,
"calibration/unique_confidence_per_question": 0.5875,
"calibration/unique_confidences": 300.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 675.0,
"completions/max_terminated_length": 675.0,
"completions/mean_length": 203.93740234375,
"completions/mean_terminated_length": 203.95697021484375,
"completions/min_length": 88.6,
"completions/min_terminated_length": 109.4,
"epoch": 0.848,
"grad_norm": 0.00827176682651043,
"learning_rate": 1e-06,
"loss": 0.0017,
"num_tokens": 899062365.0,
"reward": 1.170112943649292,
"reward_std": 0.08432346135377884,
"rewards/accuracy_reward": 0.53994140625,
"rewards/batchaccgated_coverage_0": 0.372877711057663,
"rewards/batchaccgated_coverage_1": 0.372877711057663,
"rewards/batchaccgated_coverage_10": 0.4052081048488617,
"rewards/batchaccgated_coverage_15": 0.40657660365104675,
"rewards/batchaccgated_coverage_20": 0.4111068546772003,
"rewards/batchaccgated_coverage_25": 0.4142894148826599,
"rewards/batchaccgated_coverage_5": 0.39407283067703247,
"rewards/brier_reward": 0.801174795627594,
"rewards/confidence_uniqueness_reward": 0.8652312278747558,
"rewards/format_reward": 0.99990234375,
"rewards/frontier_entropy_batch_reward": -0.44150441884994507,
"signal/accuracy_reward/centered_abs_mean": 0.066021728515625,
"signal/accuracy_reward/group_std_mean": 0.089836747944355,
"signal/accuracy_reward/group_zero_std_frac": 0.73125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.8619878768920899,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0330108642578125,
"signal/advantage_abs_mean": 0.7204357504844665,
"signal/advantage_pre_scale_abs_mean": 0.06088006719946861,
"signal/advantage_pre_scale_std": 0.12008503675460816,
"signal/advantage_std": 0.9789358139038086,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.027155644819140436,
"signal/batchaccgated_coverage_0/group_std_mean": 0.037218472361564635,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.15330489575862885,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0027155645191669466,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.027155644819140436,
"signal/batchaccgated_coverage_1/group_std_mean": 0.037218472361564635,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.15330489575862885,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0027155645191669466,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03463050127029419,
"signal/batchaccgated_coverage_10/group_std_mean": 0.050204242765903476,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.19046873450279236,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0034630501642823218,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.034565769135951996,
"signal/batchaccgated_coverage_15/group_std_mean": 0.04999654516577721,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.18985137343406677,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.003456577006727457,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.03831170275807381,
"signal/batchaccgated_coverage_20/group_std_mean": 0.055039082467556,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.21185386776924134,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.003831170406192541,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.041357473284006116,
"signal/batchaccgated_coverage_25/group_std_mean": 0.05886310636997223,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.2308925449848175,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004135747533291578,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03152990788221359,
"signal/batchaccgated_coverage_5/group_std_mean": 0.044711750000715256,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.315625,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.17752549648284913,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003152990760281682,
"signal/brier_reward/centered_abs_mean": 0.08711291253566741,
"signal/brier_reward/group_std_mean": 0.11496554613113404,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.49585837721824644,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008711291383951903,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.04876968041062355,
"signal/confidence_uniqueness_reward/group_std_mean": 0.060488971322774886,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.2833089202642441,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004876968264579773,
"signal/format_reward/centered_abs_mean": 0.000189208984375,
"signal/format_reward/group_std_mean": 0.0005524271633476019,
"signal/format_reward/group_zero_std_frac": 0.996875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.005038385838270187,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 9.46044921875e-05,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.19551956355571748,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.26076858043670653,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.1625,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.0860072255134583,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.019551957026124,
"step": 265
},
{
"calibration/aurc": 0.20964643166863067,
"calibration/batch_distribution_entropy": 0.628699236066158,
"calibration/batch_entropy_100bins": 0.7051550703304532,
"calibration/batch_entropy_10bins": 0.628699236066158,
"calibration/batch_entropy_50bins": 0.6762910242598864,
"calibration/batch_uniqueness": 0.7747866970171545,
"calibration/buffer_distribution_entropy": 0.7895420856201637,
"calibration/buffer_entropy_100bins": 0.8428426487724788,
"calibration/buffer_entropy_10bins": 0.7895420856201637,
"calibration/buffer_entropy_50bins": 0.8180176729463515,
"calibration/confidence_entropy": 0.23497960013860167,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.21305192025440314,
"calibration/coverage@15%": 0.326392031555773,
"calibration/coverage@20%": 0.547504892367906,
"calibration/coverage@25%": 0.6772099743150685,
"calibration/coverage@30%": 0.8470179488747555,
"calibration/coverage@5%": 0.11640625,
"calibration/distribution_entropy_10": 0.628699236066158,
"calibration/distribution_entropy_100": 0.7051550703304532,
"calibration/ece": 0.1956586920819896,
"calibration/mean_confidence": 0.7961682231848618,
"calibration/unique_confidence_per_question": 0.52109375,
"calibration/unique_confidences": 266.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0005859375,
"completions/max_length": 621.4,
"completions/max_terminated_length": 621.4,
"completions/mean_length": 207.3568359375,
"completions/mean_terminated_length": 207.47985229492187,
"completions/min_length": 40.0,
"completions/min_terminated_length": 103.8,
"epoch": 0.864,
"grad_norm": 0.008273676037788391,
"learning_rate": 1e-06,
"loss": -0.0001,
"num_tokens": 916172515.0,
"reward": 1.2307881832122802,
"reward_std": 0.0925519436597824,
"rewards/accuracy_reward": 0.6171875,
"rewards/batchaccgated_coverage_0": 0.3975534737110138,
"rewards/batchaccgated_coverage_1": 0.3975534737110138,
"rewards/batchaccgated_coverage_10": 0.4492712736129761,
"rewards/batchaccgated_coverage_15": 0.45667404532432554,
"rewards/batchaccgated_coverage_20": 0.46370017528533936,
"rewards/batchaccgated_coverage_25": 0.4659225523471832,
"rewards/batchaccgated_coverage_5": 0.4265682637691498,
"rewards/brier_reward": 0.8190995573997497,
"rewards/confidence_uniqueness_reward": 0.828731381893158,
"rewards/format_reward": 0.99921875,
"rewards/frontier_entropy_batch_reward": -0.47922341227531434,
"signal/accuracy_reward/centered_abs_mean": 0.07020263671875,
"signal/accuracy_reward/group_std_mean": 0.09588523507118225,
"signal/accuracy_reward/group_zero_std_frac": 0.715625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.8310398817062379,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.035101318359375,
"signal/advantage_abs_mean": 0.7335079073905945,
"signal/advantage_pre_scale_abs_mean": 0.06568295434117317,
"signal/advantage_pre_scale_std": 0.1339287281036377,
"signal/advantage_std": 0.9794105410575866,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.03260603435337543,
"signal/batchaccgated_coverage_0/group_std_mean": 0.04285504221916199,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.16704900562763214,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.003260603453963995,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.03260603435337543,
"signal/batchaccgated_coverage_1/group_std_mean": 0.04285504221916199,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.16704900562763214,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.003260603453963995,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.04576434567570686,
"signal/batchaccgated_coverage_10/group_std_mean": 0.0656816080212593,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.2350281924009323,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004576434660702944,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.048626700043678285,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06994642540812493,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.2477329283952713,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004862670227885247,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.05368155688047409,
"signal/batchaccgated_coverage_20/group_std_mean": 0.07640532702207566,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.2728260695934296,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.005368155799806118,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.055947072058916095,
"signal/batchaccgated_coverage_25/group_std_mean": 0.0791322112083435,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.28446760177612307,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.005594707280397415,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03776797130703926,
"signal/batchaccgated_coverage_5/group_std_mean": 0.05265151336789131,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.271875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.19230526983737944,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.003776797279715538,
"signal/brier_reward/centered_abs_mean": 0.08364347368478775,
"signal/brier_reward/group_std_mean": 0.1110730528831482,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.4333698809146881,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008364347275346518,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.07226071506738663,
"signal/confidence_uniqueness_reward/group_std_mean": 0.08949846178293228,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.38427419364452364,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.00722607159987092,
"signal/format_reward/centered_abs_mean": 0.00147705078125,
"signal/format_reward/group_std_mean": 0.0036875875666737556,
"signal/format_reward/group_zero_std_frac": 0.98125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.04096675962209702,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000738525390625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.20701406002044678,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2690656930208206,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.215625,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.0633200049400329,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02070140615105629,
"step": 270
},
{
"calibration/aurc": 0.3048558662437618,
"calibration/batch_distribution_entropy": 0.7760472678381269,
"calibration/batch_entropy_100bins": 0.8112185339487457,
"calibration/batch_entropy_10bins": 0.7760472678381269,
"calibration/batch_entropy_50bins": 0.7969425840935851,
"calibration/batch_uniqueness": 0.8674102783203125,
"calibration/buffer_distribution_entropy": 0.7872145257599384,
"calibration/buffer_entropy_100bins": 0.8402359861446369,
"calibration/buffer_entropy_10bins": 0.7872145257599384,
"calibration/buffer_entropy_50bins": 0.8152741284189211,
"calibration/confidence_entropy": 0.30767991330623035,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.091015625,
"calibration/coverage@15%": 0.2296875,
"calibration/coverage@20%": 0.276953125,
"calibration/coverage@25%": 0.381640625,
"calibration/coverage@30%": 0.44375,
"calibration/coverage@5%": 0.043359375,
"calibration/distribution_entropy_10": 0.7760472678381269,
"calibration/distribution_entropy_100": 0.8112185339487457,
"calibration/ece": 0.20897186590237177,
"calibration/mean_confidence": 0.6889880551556274,
"calibration/unique_confidence_per_question": 0.643359375,
"calibration/unique_confidences": 329.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 609.0,
"completions/max_terminated_length": 609.0,
"completions/mean_length": 199.37998046875,
"completions/mean_terminated_length": 199.3996154785156,
"completions/min_length": 84.6,
"completions/min_terminated_length": 102.8,
"epoch": 0.88,
"grad_norm": 0.008549376390874386,
"learning_rate": 1e-06,
"loss": -0.0004,
"num_tokens": 933361238.0,
"reward": 1.147016668319702,
"reward_std": 0.0853044331073761,
"rewards/accuracy_reward": 0.52998046875,
"rewards/batchaccgated_coverage_0": 0.3534718453884125,
"rewards/batchaccgated_coverage_1": 0.3534718453884125,
"rewards/batchaccgated_coverage_10": 0.3834411084651947,
"rewards/batchaccgated_coverage_15": 0.3852677583694458,
"rewards/batchaccgated_coverage_20": 0.39086577892303465,
"rewards/batchaccgated_coverage_25": 0.39294515252113343,
"rewards/batchaccgated_coverage_5": 0.36657083630561826,
"rewards/brier_reward": 0.7786554336547852,
"rewards/confidence_uniqueness_reward": 0.8470762014389038,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.43052550554275515,
"signal/accuracy_reward/centered_abs_mean": 0.070013427734375,
"signal/accuracy_reward/group_std_mean": 0.09500491917133332,
"signal/accuracy_reward/group_zero_std_frac": 0.71875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.2519259214401246,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0350067138671875,
"signal/advantage_abs_mean": 0.7174417853355408,
"signal/advantage_pre_scale_abs_mean": 0.0618969090282917,
"signal/advantage_pre_scale_std": 0.12277425229549407,
"signal/advantage_std": 0.9782558560371399,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.026334239915013314,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03460989110171795,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.1671230524778366,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0026334241032600404,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.026334239915013314,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03460989110171795,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.1671230524778366,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0026334241032600404,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.033635254576802255,
"signal/batchaccgated_coverage_10/group_std_mean": 0.04813460260629654,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.21496919691562652,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.003363525401800871,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03387489914894104,
"signal/batchaccgated_coverage_15/group_std_mean": 0.048321180045604706,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.21633379459381102,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0033874900080263616,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.03724322915077209,
"signal/batchaccgated_coverage_20/group_std_mean": 0.0527643121778965,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.23722701966762544,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0037243227940052747,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.03965325579047203,
"signal/batchaccgated_coverage_25/group_std_mean": 0.055769046396017076,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.25283333361148835,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.003965325560420752,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.027974789962172508,
"signal/batchaccgated_coverage_5/group_std_mean": 0.03849341571331024,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.321875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.1782465547323227,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.002797479135915637,
"signal/brier_reward/centered_abs_mean": 0.08915703296661377,
"signal/brier_reward/group_std_mean": 0.11654031276702881,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.5693645119667053,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008915703743696213,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.06098616421222687,
"signal/confidence_uniqueness_reward/group_std_mean": 0.07385204881429672,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.390488201379776,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.006098616309463978,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.011389143392443657,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.18639201521873475,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.24926036298274995,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.203125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.1875649213790893,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.018639202415943145,
"step": 275
},
{
"calibration/aurc": 0.30426706654932695,
"calibration/batch_distribution_entropy": 0.7910330076063635,
"calibration/batch_entropy_100bins": 0.8323602671671401,
"calibration/batch_entropy_10bins": 0.7910330076063635,
"calibration/batch_entropy_50bins": 0.8196394304907425,
"calibration/batch_uniqueness": 0.8951984837167505,
"calibration/buffer_distribution_entropy": 0.7850057491604059,
"calibration/buffer_entropy_100bins": 0.8397227998946711,
"calibration/buffer_entropy_10bins": 0.7850057491604059,
"calibration/buffer_entropy_50bins": 0.8148296786434728,
"calibration/confidence_entropy": 0.3147593687007963,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.14663955479452057,
"calibration/coverage@15%": 0.20488778131115462,
"calibration/coverage@20%": 0.3582329378669276,
"calibration/coverage@25%": 0.438011405332681,
"calibration/coverage@30%": 0.5267536081213307,
"calibration/coverage@5%": 0.10246605919765166,
"calibration/distribution_entropy_10": 0.7910330076063635,
"calibration/distribution_entropy_100": 0.8323602671671401,
"calibration/ece": 0.24210863760755777,
"calibration/mean_confidence": 0.6984908252595334,
"calibration/unique_confidence_per_question": 0.67578125,
"calibration/unique_confidences": 346.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 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": 692.8,
"completions/max_terminated_length": 692.8,
"completions/mean_length": 198.7703125,
"completions/mean_terminated_length": 198.86827697753907,
"completions/min_length": 18.0,
"completions/min_terminated_length": 101.6,
"epoch": 0.896,
"grad_norm": 0.00866237934678793,
"learning_rate": 1e-06,
"loss": -0.0,
"num_tokens": 950507494.0,
"reward": 1.2045636653900147,
"reward_std": 0.08532478809356689,
"rewards/accuracy_reward": 0.582421875,
"rewards/batchaccgated_coverage_0": 0.3885661602020264,
"rewards/batchaccgated_coverage_1": 0.3885661602020264,
"rewards/batchaccgated_coverage_10": 0.4258988440036774,
"rewards/batchaccgated_coverage_15": 0.42769233584403993,
"rewards/batchaccgated_coverage_20": 0.43364266157150266,
"rewards/batchaccgated_coverage_25": 0.4356557309627533,
"rewards/batchaccgated_coverage_5": 0.4098514080047607,
"rewards/brier_reward": 0.7989385843276977,
"rewards/confidence_uniqueness_reward": 0.878060233592987,
"rewards/format_reward": 0.99951171875,
"rewards/frontier_entropy_batch_reward": -0.450903856754303,
"signal/accuracy_reward/centered_abs_mean": 0.0663330078125,
"signal/accuracy_reward/group_std_mean": 0.09053425043821335,
"signal/accuracy_reward/group_zero_std_frac": 0.73125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.11031711101532,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.03316650390625,
"signal/advantage_abs_mean": 0.7152203321456909,
"signal/advantage_pre_scale_abs_mean": 0.060301771759986876,
"signal/advantage_pre_scale_std": 0.12203450351953507,
"signal/advantage_std": 0.9787332892417908,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.03338355161249638,
"signal/batchaccgated_coverage_0/group_std_mean": 0.04465327635407448,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.1849051147699356,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.003338355477899313,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.03338355161249638,
"signal/batchaccgated_coverage_1/group_std_mean": 0.04465327635407448,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.1849051147699356,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.003338355477899313,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.04311170428991318,
"signal/batchaccgated_coverage_10/group_std_mean": 0.06104911491274834,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.24437583982944489,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004311170568689704,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.04379637986421585,
"signal/batchaccgated_coverage_15/group_std_mean": 0.06191677451133728,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.24777641594409944,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004379638191312552,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.0473410002887249,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06694948822259902,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.26739788949489596,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004734100308269262,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.047765230387449266,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06732234954833985,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.2751908332109451,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004776523355394602,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03754683248698711,
"signal/batchaccgated_coverage_5/group_std_mean": 0.05164859890937805,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.21125857532024384,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0037546833511441944,
"signal/brier_reward/centered_abs_mean": 0.08105404824018478,
"signal/brier_reward/group_std_mean": 0.10782659500837326,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.49806665182113646,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008105404675006866,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.04076760113239288,
"signal/confidence_uniqueness_reward/group_std_mean": 0.0522532619535923,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.25780007541179656,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004076760355383157,
"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": 0.024459636956453323,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0004730224609375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.19032479226589202,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2517787963151932,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.18125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.1073362946510314,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.019032479822635652,
"step": 280
},
{
"calibration/aurc": 0.3027295877757971,
"calibration/batch_distribution_entropy": 0.7428323064984742,
"calibration/batch_entropy_100bins": 0.777480571794617,
"calibration/batch_entropy_10bins": 0.7428323064984742,
"calibration/batch_entropy_50bins": 0.7560963454953145,
"calibration/batch_uniqueness": 0.8409815461164243,
"calibration/buffer_distribution_entropy": 0.7788881332701182,
"calibration/buffer_entropy_100bins": 0.8364284271011246,
"calibration/buffer_entropy_10bins": 0.7788881332701182,
"calibration/buffer_entropy_50bins": 0.8109678427511046,
"calibration/confidence_entropy": 0.2835003330500979,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0390625,
"calibration/coverage@15%": 0.277734375,
"calibration/coverage@20%": 0.370703125,
"calibration/coverage@25%": 0.5015625,
"calibration/coverage@30%": 0.558203125,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.7428323064984742,
"calibration/distribution_entropy_100": 0.777480571794617,
"calibration/ece": 0.237708429762388,
"calibration/mean_confidence": 0.7350196133105451,
"calibration/unique_confidence_per_question": 0.55859375,
"calibration/unique_confidences": 286.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00029296875,
"completions/max_length": 615.6,
"completions/max_terminated_length": 615.6,
"completions/mean_length": 194.44111328125,
"completions/mean_terminated_length": 194.49887084960938,
"completions/min_length": 61.0,
"completions/min_terminated_length": 101.8,
"epoch": 0.912,
"grad_norm": 0.008404405787587166,
"learning_rate": 1e-06,
"loss": -0.0011,
"num_tokens": 967549867.0,
"reward": 1.1999149560928344,
"reward_std": 0.08682571947574616,
"rewards/accuracy_reward": 0.573046875,
"rewards/batchaccgated_coverage_0": 0.3957234680652618,
"rewards/batchaccgated_coverage_1": 0.3957234680652618,
"rewards/batchaccgated_coverage_10": 0.43918827176094055,
"rewards/batchaccgated_coverage_15": 0.4421947062015533,
"rewards/batchaccgated_coverage_20": 0.449607789516449,
"rewards/batchaccgated_coverage_25": 0.45175438523292544,
"rewards/batchaccgated_coverage_5": 0.4309255242347717,
"rewards/brier_reward": 0.7931956052780151,
"rewards/confidence_uniqueness_reward": 0.8119370102882385,
"rewards/format_reward": 0.99970703125,
"rewards/frontier_entropy_batch_reward": -0.4748704552650452,
"signal/accuracy_reward/centered_abs_mean": 0.081494140625,
"signal/accuracy_reward/group_std_mean": 0.10337824821472168,
"signal/accuracy_reward/group_zero_std_frac": 0.715625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 3.149342250823975,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0407470703125,
"signal/advantage_abs_mean": 0.71235111951828,
"signal/advantage_pre_scale_abs_mean": 0.06253414377570152,
"signal/advantage_pre_scale_std": 0.1261533334851265,
"signal/advantage_std": 0.9767374992370605,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.02390987053513527,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03350846357643604,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.18195807933807373,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0023909871699288486,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.02390987053513527,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03350846357643604,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.18195807933807373,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0023909871699288486,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03142364993691445,
"signal/batchaccgated_coverage_10/group_std_mean": 0.04850826933979988,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.24394859373569489,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.003142365114763379,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03292708955705166,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05053712651133537,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.2522888481616974,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0032927090302109717,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.03887040838599205,
"signal/batchaccgated_coverage_20/group_std_mean": 0.05821322351694107,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.30077292323112487,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0038870408199727534,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04064108356833458,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06018582060933113,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.3140357851982117,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004064108338207006,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.0289675347507,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04450409859418869,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.29375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.22409008145332338,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.00289675360545516,
"signal/brier_reward/centered_abs_mean": 0.08763111978769303,
"signal/brier_reward/group_std_mean": 0.11560369431972503,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.7003116548061371,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008763112127780914,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.06957556307315826,
"signal/confidence_uniqueness_reward/group_std_mean": 0.08646434098482132,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.5491433203220367,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.006957556493580342,
"signal/format_reward/centered_abs_mean": 0.000567626953125,
"signal/format_reward/group_std_mean": 0.0016572814900428056,
"signal/format_reward/group_zero_std_frac": 0.990625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.02477443516254425,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0002838134765625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.18756958842277527,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.25171841084957125,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.20625,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.4943580150604248,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.018756958469748498,
"step": 285
},
{
"calibration/aurc": 0.32839787434682044,
"calibration/batch_distribution_entropy": 0.801611021029256,
"calibration/batch_entropy_100bins": 0.8445312389157094,
"calibration/batch_entropy_10bins": 0.801611021029256,
"calibration/batch_entropy_50bins": 0.831691516488806,
"calibration/batch_uniqueness": 0.9009028167369906,
"calibration/buffer_distribution_entropy": 0.7778570457571966,
"calibration/buffer_entropy_100bins": 0.8356058351223202,
"calibration/buffer_entropy_10bins": 0.7778570457571966,
"calibration/buffer_entropy_50bins": 0.8099420674688329,
"calibration/confidence_entropy": 0.32253098998543256,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.118359375,
"calibration/coverage@15%": 0.21653543909001954,
"calibration/coverage@20%": 0.2556224009295499,
"calibration/coverage@25%": 0.34196810787671234,
"calibration/coverage@30%": 0.42284506482387474,
"calibration/coverage@5%": 0.075,
"calibration/distribution_entropy_10": 0.801611021029256,
"calibration/distribution_entropy_100": 0.8445312389157094,
"calibration/ece": 0.24593697343972543,
"calibration/mean_confidence": 0.6854968666941871,
"calibration/unique_confidence_per_question": 0.609375,
"calibration/unique_confidences": 312.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00029296875,
"completions/max_length": 744.4,
"completions/max_terminated_length": 744.4,
"completions/mean_length": 191.9837890625,
"completions/mean_terminated_length": 192.04074401855468,
"completions/min_length": 59.2,
"completions/min_terminated_length": 99.2,
"epoch": 0.928,
"grad_norm": 0.01045648381114006,
"learning_rate": 1e-06,
"loss": 0.001,
"num_tokens": 984542597.0,
"reward": 1.1734341144561768,
"reward_std": 0.07837996929883957,
"rewards/accuracy_reward": 0.54404296875,
"rewards/batchaccgated_coverage_0": 0.36571348905563356,
"rewards/batchaccgated_coverage_1": 0.36571348905563356,
"rewards/batchaccgated_coverage_10": 0.4118075966835022,
"rewards/batchaccgated_coverage_15": 0.41431081891059873,
"rewards/batchaccgated_coverage_20": 0.4196390092372894,
"rewards/batchaccgated_coverage_25": 0.42054404616355895,
"rewards/batchaccgated_coverage_5": 0.3983059227466583,
"rewards/brier_reward": 0.8005460619926452,
"rewards/confidence_uniqueness_reward": 0.8736016988754273,
"rewards/format_reward": 0.99970703125,
"rewards/frontier_entropy_batch_reward": -0.4545911192893982,
"signal/accuracy_reward/centered_abs_mean": 0.056805419921875,
"signal/accuracy_reward/group_std_mean": 0.07906077951192855,
"signal/accuracy_reward/group_zero_std_frac": 0.759375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.7254481315612793,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0284027099609375,
"signal/advantage_abs_mean": 0.7162193536758423,
"signal/advantage_pre_scale_abs_mean": 0.054822988063097,
"signal/advantage_pre_scale_std": 0.11606567800045013,
"signal/advantage_std": 0.9784825801849365,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.02752493880689144,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03668419010937214,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.35,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.16997956931591035,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0027524939738214015,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.02752493880689144,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03668419010937214,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.35,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.16997956931591035,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0027524939738214015,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03615510016679764,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05362729802727699,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.35,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.22355604469776152,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0036155101377516987,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.037400002032518385,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05544038340449333,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.35,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.23236773610115052,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0037400003988295794,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.039513856172561646,
"signal/batchaccgated_coverage_20/group_std_mean": 0.05852712020277977,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.35,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.244587305188179,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.003951385710388422,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.040053793042898175,
"signal/batchaccgated_coverage_25/group_std_mean": 0.05918071568012238,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.35,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.24787524938583375,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004005379509180784,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03267774805426597,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04768794849514961,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.35,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.20389316380023956,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0032677749171853066,
"signal/brier_reward/centered_abs_mean": 0.08034662157297134,
"signal/brier_reward/group_std_mean": 0.10586177706718444,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.48977477550506593,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008034662622958422,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.04792383387684822,
"signal/confidence_uniqueness_reward/group_std_mean": 0.059908661246299746,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.2962785542011261,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004792383406311274,
"signal/format_reward/centered_abs_mean": 0.000567626953125,
"signal/format_reward/group_std_mean": 0.0016572814900428056,
"signal/format_reward/group_zero_std_frac": 0.990625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.016047720983624458,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0002838134765625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.19745397567749023,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.25981183648109435,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.178125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.205248522758484,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.019745399430394172,
"step": 290
},
{
"calibration/aurc": 0.23352131547418803,
"calibration/batch_distribution_entropy": 0.8063181281576497,
"calibration/batch_entropy_100bins": 0.8241308460252492,
"calibration/batch_entropy_10bins": 0.8063181281576497,
"calibration/batch_entropy_50bins": 0.8123251280031063,
"calibration/batch_uniqueness": 0.8702497617722766,
"calibration/buffer_distribution_entropy": 0.7821345510747968,
"calibration/buffer_entropy_100bins": 0.8385322702451419,
"calibration/buffer_entropy_10bins": 0.7821345510747968,
"calibration/buffer_entropy_50bins": 0.8133301294982619,
"calibration/confidence_entropy": 0.32030045714573097,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.21196260396281802,
"calibration/coverage@15%": 0.38756682938586395,
"calibration/coverage@20%": 0.5047075371244387,
"calibration/coverage@25%": 0.5963289373968765,
"calibration/coverage@30%": 0.662120835491731,
"calibration/coverage@5%": 0.10949960249510764,
"calibration/distribution_entropy_10": 0.8063181281576497,
"calibration/distribution_entropy_100": 0.8241308460252492,
"calibration/ece": 0.16154086667174433,
"calibration/mean_confidence": 0.6687827685498403,
"calibration/unique_confidence_per_question": 0.606640625,
"calibration/unique_confidences": 310.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00048828125,
"completions/max_length": 770.4,
"completions/max_terminated_length": 770.4,
"completions/mean_length": 185.7771484375,
"completions/mean_terminated_length": 185.86882934570312,
"completions/min_length": 18.0,
"completions/min_terminated_length": 94.6,
"epoch": 0.944,
"grad_norm": 0.008868498727679253,
"learning_rate": 1e-06,
"loss": 0.0033,
"num_tokens": 1001420379.0,
"reward": 1.165120816230774,
"reward_std": 0.09554860591888428,
"rewards/accuracy_reward": 0.550390625,
"rewards/batchaccgated_coverage_0": 0.3494316875934601,
"rewards/batchaccgated_coverage_1": 0.3494316875934601,
"rewards/batchaccgated_coverage_10": 0.3904098868370056,
"rewards/batchaccgated_coverage_15": 0.3922972917556763,
"rewards/batchaccgated_coverage_20": 0.3939199447631836,
"rewards/batchaccgated_coverage_25": 0.39588758945465086,
"rewards/batchaccgated_coverage_5": 0.377797394990921,
"rewards/brier_reward": 0.8024485945701599,
"rewards/confidence_uniqueness_reward": 0.8666816473007202,
"rewards/format_reward": 0.99736328125,
"rewards/frontier_entropy_batch_reward": -0.4058672308921814,
"signal/accuracy_reward/centered_abs_mean": 0.0781982421875,
"signal/accuracy_reward/group_std_mean": 0.10564855337142945,
"signal/accuracy_reward/group_zero_std_frac": 0.68125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.1944111585617065,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.03909912109375,
"signal/advantage_abs_mean": 0.695769476890564,
"signal/advantage_pre_scale_abs_mean": 0.06632322520017624,
"signal/advantage_pre_scale_std": 0.13217144310474396,
"signal/advantage_std": 0.9788856506347656,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.030407553166151048,
"signal/batchaccgated_coverage_0/group_std_mean": 0.04082559272646904,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.16954272240400314,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.003040755447000265,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.030407553166151048,
"signal/batchaccgated_coverage_1/group_std_mean": 0.04082559272646904,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.16954272240400314,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.003040755447000265,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03835934065282345,
"signal/batchaccgated_coverage_10/group_std_mean": 0.054130587726831436,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.21505818665027618,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0038359342142939568,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03937883675098419,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05551670640707016,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.22074773907661438,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.003937883814796805,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.040260034054517745,
"signal/batchaccgated_coverage_20/group_std_mean": 0.05662956014275551,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.22638989090919495,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0040260035078972575,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04125453159213066,
"signal/batchaccgated_coverage_25/group_std_mean": 0.057932039350271226,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.2316810131072998,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004125453252345324,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03496180698275566,
"signal/batchaccgated_coverage_5/group_std_mean": 0.048563698679208754,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.30625,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.19406530261039734,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0034961807541549206,
"signal/brier_reward/centered_abs_mean": 0.08858360201120377,
"signal/brier_reward/group_std_mean": 0.11632589846849442,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.49475149512290956,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008858360722661019,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.057854165136814115,
"signal/confidence_uniqueness_reward/group_std_mean": 0.0772600993514061,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.3443441689014435,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0057854167185723785,
"signal/format_reward/centered_abs_mean": 0.005096435546875,
"signal/format_reward/group_std_mean": 0.014579224260523915,
"signal/format_reward/group_zero_std_frac": 0.91875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.14164469838142396,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0025482177734375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2037956655025482,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2710902035236359,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.121875,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.1580429911613463,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0203795675188303,
"step": 295
},
{
"calibration/aurc": 0.27850018044820013,
"calibration/batch_distribution_entropy": 0.6736399917321533,
"calibration/batch_entropy_100bins": 0.7277351463199947,
"calibration/batch_entropy_10bins": 0.6736399917321533,
"calibration/batch_entropy_50bins": 0.7026574473085763,
"calibration/batch_uniqueness": 0.7993698988296419,
"calibration/buffer_distribution_entropy": 0.7811231080100745,
"calibration/buffer_entropy_100bins": 0.8363335730487644,
"calibration/buffer_entropy_10bins": 0.7811231080100745,
"calibration/buffer_entropy_50bins": 0.811019603033816,
"calibration/confidence_entropy": 0.2494676052939062,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.18359375,
"calibration/coverage@20%": 0.2504792991682975,
"calibration/coverage@25%": 0.5293855491682974,
"calibration/coverage@30%": 0.6748394691780822,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.6736399917321533,
"calibration/distribution_entropy_100": 0.7277351463199947,
"calibration/ece": 0.2451664995753689,
"calibration/mean_confidence": 0.7520032391029527,
"calibration/unique_confidence_per_question": 0.482421875,
"calibration/unique_confidences": 247.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 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": 627.4,
"completions/max_terminated_length": 627.4,
"completions/mean_length": 182.953515625,
"completions/mean_terminated_length": 183.07962036132812,
"completions/min_length": 36.2,
"completions/min_terminated_length": 92.6,
"epoch": 0.96,
"grad_norm": 0.008852118626236916,
"learning_rate": 1e-06,
"loss": -0.0053,
"num_tokens": 1018234143.0,
"reward": 1.1815873384475708,
"reward_std": 0.07041721493005752,
"rewards/accuracy_reward": 0.5435546875,
"rewards/batchaccgated_coverage_0": 0.4017432153224945,
"rewards/batchaccgated_coverage_1": 0.4017432153224945,
"rewards/batchaccgated_coverage_10": 0.4297162055969238,
"rewards/batchaccgated_coverage_15": 0.43861175179481504,
"rewards/batchaccgated_coverage_20": 0.44260507822036743,
"rewards/batchaccgated_coverage_25": 0.4428340792655945,
"rewards/batchaccgated_coverage_5": 0.41624986529350283,
"rewards/brier_reward": 0.7879561305046081,
"rewards/confidence_uniqueness_reward": 0.8181966185569763,
"rewards/format_reward": 0.998828125,
"rewards/frontier_entropy_batch_reward": -0.47569699883461,
"signal/accuracy_reward/centered_abs_mean": 0.0516845703125,
"signal/accuracy_reward/group_std_mean": 0.0751104511320591,
"signal/accuracy_reward/group_zero_std_frac": 0.7625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.8486485958099366,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.02584228515625,
"signal/advantage_abs_mean": 0.7043779611587524,
"signal/advantage_pre_scale_abs_mean": 0.04744292572140694,
"signal/advantage_pre_scale_std": 0.10692166537046432,
"signal/advantage_std": 0.9775358080863953,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.021797166764736177,
"signal/batchaccgated_coverage_0/group_std_mean": 0.029483487457036973,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.15394878685474395,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0021797166904434563,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.021797166764736177,
"signal/batchaccgated_coverage_1/group_std_mean": 0.029483487457036973,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.15394878685474395,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0021797166904434563,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.027605806849896906,
"signal/batchaccgated_coverage_10/group_std_mean": 0.03990558162331581,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.19526177793741226,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0027605806477367877,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.03187102824449539,
"signal/batchaccgated_coverage_15/group_std_mean": 0.04629060104489326,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.22664715349674225,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0031871029641479254,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.03498369306325912,
"signal/batchaccgated_coverage_20/group_std_mean": 0.05042875334620476,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.24569307267665863,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.0034983694087713955,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.03452742919325828,
"signal/batchaccgated_coverage_25/group_std_mean": 0.049974224716424945,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.24355627000331878,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.003452743077650666,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.023767183534801006,
"signal/batchaccgated_coverage_5/group_std_mean": 0.0333886593580246,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.16855373084545136,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0023767184698954224,
"signal/brier_reward/centered_abs_mean": 0.07507161647081376,
"signal/brier_reward/group_std_mean": 0.10052584856748581,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.5353548109531403,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.00750716207548976,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.07436529397964478,
"signal/confidence_uniqueness_reward/group_std_mean": 0.09100949019193649,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.5432931303977966,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.007436529546976089,
"signal/format_reward/centered_abs_mean": 0.00225830078125,
"signal/format_reward/group_std_mean": 0.0062928175088018175,
"signal/format_reward/group_zero_std_frac": 0.965625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.09227947928011418,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.001129150390625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.17627755105495452,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.23899804651737214,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.203125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.2495132565498352,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.017627755366265773,
"step": 300
},
{
"epoch": 0.96,
"eval_completions/clipped_ratio": 0.0,
"eval_completions/max_length": 378.75,
"eval_completions/max_terminated_length": 378.75,
"eval_completions/mean_length": 183.39102935791016,
"eval_completions/mean_terminated_length": 183.39102935791016,
"eval_completions/min_length": 109.25,
"eval_completions/min_terminated_length": 109.25,
"eval_loss": 0.0,
"eval_num_tokens": 1018234143.0,
"eval_reward": 0.8031385093927383,
"eval_reward_std": 0.24449624493718147,
"eval_rewards/accuracy_reward": 0.443359375,
"eval_rewards/batchaccgated_coverage_0": 0.03105087112635374,
"eval_rewards/batchaccgated_coverage_1": 0.03105087112635374,
"eval_rewards/batchaccgated_coverage_10": 0.030407235026359558,
"eval_rewards/batchaccgated_coverage_15": 0.02986953780055046,
"eval_rewards/batchaccgated_coverage_20": 0.02712287474423647,
"eval_rewards/batchaccgated_coverage_25": 0.02441434096544981,
"eval_rewards/batchaccgated_coverage_5": 0.03105087112635374,
"eval_rewards/brier_reward": 0.7583034485578537,
"eval_rewards/confidence_uniqueness_reward": 0.851318359375,
"eval_rewards/format_reward": 1.0,
"eval_rewards/frontier_entropy_batch_reward": -1.0,
"eval_runtime": 19.3317,
"eval_samples_per_second": 25.864,
"eval_signal/accuracy_reward/centered_abs_mean": 0.4761962890625,
"eval_signal/accuracy_reward/group_std_mean": 0.4952752813696861,
"eval_signal/accuracy_reward/group_zero_std_frac": 0.0,
"eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9745659232139587,
"eval_signal/accuracy_reward/weight": 0.5,
"eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.23809814453125,
"eval_signal/advantage_abs_mean": 0.9218599498271942,
"eval_signal/advantage_pre_scale_abs_mean": 0.22571881115436554,
"eval_signal/advantage_pre_scale_std": 0.2417391650378704,
"eval_signal/advantage_std": 0.9877137243747711,
"eval_signal/batchaccgated_coverage_0/centered_abs_mean": 0.05186370946466923,
"eval_signal/batchaccgated_coverage_0/group_std_mean": 0.06709158234298229,
"eval_signal/batchaccgated_coverage_0/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.021218573674559593,
"eval_signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.005186370923183858,
"eval_signal/batchaccgated_coverage_1/centered_abs_mean": 0.05186370946466923,
"eval_signal/batchaccgated_coverage_1/group_std_mean": 0.06709158234298229,
"eval_signal/batchaccgated_coverage_1/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.021218573674559593,
"eval_signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.005186370923183858,
"eval_signal/batchaccgated_coverage_10/centered_abs_mean": 0.05080135352909565,
"eval_signal/batchaccgated_coverage_10/group_std_mean": 0.06580364890396595,
"eval_signal/batchaccgated_coverage_10/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.020786314737051725,
"eval_signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.005080135422758758,
"eval_signal/batchaccgated_coverage_15/centered_abs_mean": 0.04963265731930733,
"eval_signal/batchaccgated_coverage_15/group_std_mean": 0.06431423034518957,
"eval_signal/batchaccgated_coverage_15/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.02030557207763195,
"eval_signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.004963265964761376,
"eval_signal/batchaccgated_coverage_20/centered_abs_mean": 0.0435302359983325,
"eval_signal/batchaccgated_coverage_20/group_std_mean": 0.05719771608710289,
"eval_signal/batchaccgated_coverage_20/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.01782950502820313,
"eval_signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.00435302359983325,
"eval_signal/batchaccgated_coverage_25/centered_abs_mean": 0.03706817701458931,
"eval_signal/batchaccgated_coverage_25/group_std_mean": 0.04884832352399826,
"eval_signal/batchaccgated_coverage_25/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.015171174658462405,
"eval_signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.0037068177480250597,
"eval_signal/batchaccgated_coverage_5/centered_abs_mean": 0.05186370946466923,
"eval_signal/batchaccgated_coverage_5/group_std_mean": 0.06709158234298229,
"eval_signal/batchaccgated_coverage_5/group_zero_std_frac": 0.0,
"eval_signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.021218573674559593,
"eval_signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"eval_signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.005186370923183858,
"eval_signal/brier_reward/centered_abs_mean": 0.2845764011144638,
"eval_signal/brier_reward/group_std_mean": 0.34831059724092484,
"eval_signal/brier_reward/group_zero_std_frac": 0.0,
"eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1165114063769579,
"eval_signal/brier_reward/weight": 0.10000000149011612,
"eval_signal/brier_reward/weighted_centered_abs_mean": 0.028457640204578638,
"eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.0831298828125,
"eval_signal/confidence_uniqueness_reward/group_std_mean": 0.10036689788103104,
"eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.03387655084952712,
"eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.008312988560646772,
"eval_signal/format_reward/centered_abs_mean": 0.0,
"eval_signal/format_reward/group_std_mean": 0.0,
"eval_signal/format_reward/group_zero_std_frac": 1.0,
"eval_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"eval_signal/format_reward/weight": 0.5,
"eval_signal/format_reward/weighted_centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"eval_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"eval_steps_per_second": 0.207,
"step": 300
},
{
"epoch": 0.96,
"step": 300,
"train_probe_completions/clipped_ratio": 0.0,
"train_probe_completions/max_length": 469.5,
"train_probe_completions/max_terminated_length": 469.5,
"train_probe_completions/mean_length": 178.6131477355957,
"train_probe_completions/mean_terminated_length": 178.6131477355957,
"train_probe_completions/min_length": 101.5,
"train_probe_completions/min_terminated_length": 101.5,
"train_probe_loss": 0.0,
"train_probe_num_tokens": 1018234143.0,
"train_probe_reward": 0.9301466345787048,
"train_probe_reward_std": 0.24371350556612015,
"train_probe_rewards/accuracy_reward": 0.642578125,
"train_probe_rewards/batchaccgated_coverage_0": 0.04586124513298273,
"train_probe_rewards/batchaccgated_coverage_1": 0.04586124513298273,
"train_probe_rewards/batchaccgated_coverage_10": 0.038542845752090216,
"train_probe_rewards/batchaccgated_coverage_15": 0.05498477537184954,
"train_probe_rewards/batchaccgated_coverage_20": 0.07146968087181449,
"train_probe_rewards/batchaccgated_coverage_25": 0.10823106486350298,
"train_probe_rewards/batchaccgated_coverage_5": 0.045582485385239124,
"train_probe_rewards/brier_reward": 0.8579739630222321,
"train_probe_rewards/confidence_uniqueness_reward": 0.820068359375,
"train_probe_rewards/format_reward": 1.0,
"train_probe_rewards/frontier_entropy_batch_reward": -1.0,
"train_probe_runtime": 21.4662,
"train_probe_samples_per_second": 23.292,
"train_probe_signal/accuracy_reward/centered_abs_mean": 0.4505615234375,
"train_probe_signal/accuracy_reward/group_std_mean": 0.4818578064441681,
"train_probe_signal/accuracy_reward/group_zero_std_frac": 0.0,
"train_probe_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.926200270652771,
"train_probe_signal/accuracy_reward/weight": 0.5,
"train_probe_signal/accuracy_reward/weighted_centered_abs_mean": 0.22528076171875,
"train_probe_signal/advantage_abs_mean": 0.8939660042524338,
"train_probe_signal/advantage_pre_scale_abs_mean": 0.21787172555923462,
"train_probe_signal/advantage_pre_scale_std": 0.2409588247537613,
"train_probe_signal/advantage_std": 0.9877124428749084,
"train_probe_signal/batchaccgated_coverage_0/centered_abs_mean": 0.09329593554139137,
"train_probe_signal/batchaccgated_coverage_0/group_std_mean": 0.12869727611541748,
"train_probe_signal/batchaccgated_coverage_0/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.038347952999174595,
"train_probe_signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.009329593740403652,
"train_probe_signal/batchaccgated_coverage_1/centered_abs_mean": 0.09329593554139137,
"train_probe_signal/batchaccgated_coverage_1/group_std_mean": 0.12869727611541748,
"train_probe_signal/batchaccgated_coverage_1/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.038347952999174595,
"train_probe_signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.009329593740403652,
"train_probe_signal/batchaccgated_coverage_10/centered_abs_mean": 0.06261359062045813,
"train_probe_signal/batchaccgated_coverage_10/group_std_mean": 0.09102711454033852,
"train_probe_signal/batchaccgated_coverage_10/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.025796424597501755,
"train_probe_signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.006261359318159521,
"train_probe_signal/batchaccgated_coverage_15/centered_abs_mean": 0.06433605030179024,
"train_probe_signal/batchaccgated_coverage_15/group_std_mean": 0.08675757050514221,
"train_probe_signal/batchaccgated_coverage_15/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.02633900847285986,
"train_probe_signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.006433605216443539,
"train_probe_signal/batchaccgated_coverage_20/centered_abs_mean": 0.0728608425706625,
"train_probe_signal/batchaccgated_coverage_20/group_std_mean": 0.09130198135972023,
"train_probe_signal/batchaccgated_coverage_20/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.02973267761990428,
"train_probe_signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.007286084408406168,
"train_probe_signal/batchaccgated_coverage_25/centered_abs_mean": 0.11567197926342487,
"train_probe_signal/batchaccgated_coverage_25/group_std_mean": 0.13595649879425764,
"train_probe_signal/batchaccgated_coverage_25/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.04738030210137367,
"train_probe_signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.011567198089323938,
"train_probe_signal/batchaccgated_coverage_5/centered_abs_mean": 0.09171407669782639,
"train_probe_signal/batchaccgated_coverage_5/group_std_mean": 0.12655112892389297,
"train_probe_signal/batchaccgated_coverage_5/group_zero_std_frac": 0.0,
"train_probe_signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.03771956078708172,
"train_probe_signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"train_probe_signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.009171408135443926,
"train_probe_signal/brier_reward/centered_abs_mean": 0.18815095350146294,
"train_probe_signal/brier_reward/group_std_mean": 0.27081988006830215,
"train_probe_signal/brier_reward/group_zero_std_frac": 0.0,
"train_probe_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.07744615990668535,
"train_probe_signal/brier_reward/weight": 0.10000000149011612,
"train_probe_signal/brier_reward/weighted_centered_abs_mean": 0.01881509553641081,
"train_probe_signal/confidence_uniqueness_reward/centered_abs_mean": 0.1126556396484375,
"train_probe_signal/confidence_uniqueness_reward/group_std_mean": 0.13106262311339378,
"train_probe_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"train_probe_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.04636909253895283,
"train_probe_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"train_probe_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.011265564477071166,
"train_probe_signal/format_reward/centered_abs_mean": 0.0,
"train_probe_signal/format_reward/group_std_mean": 0.0,
"train_probe_signal/format_reward/group_zero_std_frac": 1.0,
"train_probe_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/format_reward/weight": 0.5,
"train_probe_signal/format_reward/weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"train_probe_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"train_probe_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"train_probe_steps_per_second": 0.186
},
{
"calibration/aurc": 0.23775545366658327,
"calibration/batch_distribution_entropy": 0.7648731437289149,
"calibration/batch_entropy_100bins": 0.8197225415680605,
"calibration/batch_entropy_10bins": 0.7648731437289149,
"calibration/batch_entropy_50bins": 0.8060763000463,
"calibration/batch_uniqueness": 0.8816739549823837,
"calibration/buffer_distribution_entropy": 0.7795439475190415,
"calibration/buffer_entropy_100bins": 0.8353093558775984,
"calibration/buffer_entropy_10bins": 0.7795439475190415,
"calibration/buffer_entropy_50bins": 0.8099060891142141,
"calibration/confidence_entropy": 0.3097802994266544,
"calibration/coverage@0%": 0.053515625,
"calibration/coverage@1%": 0.06171875,
"calibration/coverage@10%": 0.26328125,
"calibration/coverage@15%": 0.459375,
"calibration/coverage@20%": 0.52578125,
"calibration/coverage@25%": 0.589453125,
"calibration/coverage@30%": 0.684375,
"calibration/coverage@5%": 0.080859375,
"calibration/distribution_entropy_10": 0.7648731437289149,
"calibration/distribution_entropy_100": 0.8197225415680605,
"calibration/ece": 0.16090970729464193,
"calibration/mean_confidence": 0.7144808733325128,
"calibration/unique_confidence_per_question": 0.63046875,
"calibration/unique_confidences": 322.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00078125,
"completions/max_length": 834.0,
"completions/max_terminated_length": 834.0,
"completions/mean_length": 187.10166015625,
"completions/mean_terminated_length": 187.24796447753906,
"completions/min_length": 18.4,
"completions/min_terminated_length": 95.0,
"epoch": 0.976,
"grad_norm": 0.00896392110735178,
"learning_rate": 1e-06,
"loss": -0.0029,
"num_tokens": 1035011184.0,
"reward": 1.1877195119857789,
"reward_std": 0.08008053749799729,
"rewards/accuracy_reward": 0.562890625,
"rewards/batchaccgated_coverage_0": 0.37995734214782717,
"rewards/batchaccgated_coverage_1": 0.37995734214782717,
"rewards/batchaccgated_coverage_10": 0.4141102731227875,
"rewards/batchaccgated_coverage_15": 0.417661315202713,
"rewards/batchaccgated_coverage_20": 0.41931965947151184,
"rewards/batchaccgated_coverage_25": 0.42064470052719116,
"rewards/batchaccgated_coverage_5": 0.4053920328617096,
"rewards/brier_reward": 0.7955968737602234,
"rewards/confidence_uniqueness_reward": 0.8782744407653809,
"rewards/format_reward": 0.99921875,
"rewards/frontier_entropy_batch_reward": -0.444265753030777,
"signal/accuracy_reward/centered_abs_mean": 0.06607666015625,
"signal/accuracy_reward/group_std_mean": 0.09089337512850762,
"signal/accuracy_reward/group_zero_std_frac": 0.725,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.6580353736877442,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.033038330078125,
"signal/advantage_abs_mean": 0.7119146227836609,
"signal/advantage_pre_scale_abs_mean": 0.057272279262542726,
"signal/advantage_pre_scale_std": 0.11449765861034393,
"signal/advantage_std": 0.9791268706321716,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.02979014813899994,
"signal/batchaccgated_coverage_0/group_std_mean": 0.03993955552577973,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.15968257784843445,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0029790148138999937,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.02979014813899994,
"signal/batchaccgated_coverage_1/group_std_mean": 0.03993955552577973,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.15968257784843445,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0029790148138999937,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.03748157247900963,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05240300595760346,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.20805883556604385,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0037481572944670916,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.038313575088977814,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05379014909267425,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.21264007836580276,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0038313576951622963,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.039554671943187715,
"signal/batchaccgated_coverage_20/group_std_mean": 0.05549765974283218,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.22301185727119446,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.003955467417836189,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.039445796981453896,
"signal/batchaccgated_coverage_25/group_std_mean": 0.05538865253329277,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.22597005665302278,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.00394457969814539,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03451544865965843,
"signal/batchaccgated_coverage_5/group_std_mean": 0.047826632112264636,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.1880588412284851,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0034515449311584233,
"signal/brier_reward/centered_abs_mean": 0.07830462828278542,
"signal/brier_reward/group_std_mean": 0.10303362607955932,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.4172621190547943,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.007830462791025639,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.04654725193977356,
"signal/confidence_uniqueness_reward/group_std_mean": 0.05877678170800209,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.28888666480779646,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004654725361615419,
"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": 0.03695552423596382,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.00074462890625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.19571826457977295,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.26424733400344846,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.15625,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.061959934234619,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.019571825861930847,
"step": 305
},
{
"calibration/aurc": 0.3482168967959146,
"calibration/batch_distribution_entropy": 0.7580045524353574,
"calibration/batch_entropy_100bins": 0.8089741573269077,
"calibration/batch_entropy_10bins": 0.7580045524353574,
"calibration/batch_entropy_50bins": 0.7911056346189158,
"calibration/batch_uniqueness": 0.8742313874653407,
"calibration/buffer_distribution_entropy": 0.775543393968797,
"calibration/buffer_entropy_100bins": 0.8330033463588047,
"calibration/buffer_entropy_10bins": 0.775543393968797,
"calibration/buffer_entropy_50bins": 0.8071748465412876,
"calibration/confidence_entropy": 0.30808443560542237,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.03046875,
"calibration/coverage@15%": 0.09140625,
"calibration/coverage@20%": 0.16561425589390963,
"calibration/coverage@25%": 0.2836171856193264,
"calibration/coverage@30%": 0.44721529574865854,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.7580045524353574,
"calibration/distribution_entropy_100": 0.8089741573269077,
"calibration/ece": 0.28718872964131403,
"calibration/mean_confidence": 0.7262240215273776,
"calibration/unique_confidence_per_question": 0.60546875,
"calibration/unique_confidences": 310.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00302734375,
"completions/max_length": 1010.4,
"completions/max_terminated_length": 1010.4,
"completions/mean_length": 191.791796875,
"completions/mean_terminated_length": 192.37608642578124,
"completions/min_length": 0.0,
"completions/min_terminated_length": 92.6,
"epoch": 0.992,
"grad_norm": 0.015328600071370602,
"learning_rate": 1e-06,
"loss": -0.0178,
"num_tokens": 1052103612.0,
"reward": 1.1663033723831178,
"reward_std": 0.09818619191646576,
"rewards/accuracy_reward": 0.54892578125,
"rewards/batchaccgated_coverage_0": 0.35357803106307983,
"rewards/batchaccgated_coverage_1": 0.35357803106307983,
"rewards/batchaccgated_coverage_10": 0.413808262348175,
"rewards/batchaccgated_coverage_15": 0.4213964521884918,
"rewards/batchaccgated_coverage_20": 0.42779074907302855,
"rewards/batchaccgated_coverage_25": 0.4300235152244568,
"rewards/batchaccgated_coverage_5": 0.3994297027587891,
"rewards/brier_reward": 0.7722519993782043,
"rewards/confidence_uniqueness_reward": 0.8389286041259766,
"rewards/format_reward": 0.99638671875,
"rewards/frontier_entropy_batch_reward": -0.47431432008743285,
"signal/accuracy_reward/centered_abs_mean": 0.067327880859375,
"signal/accuracy_reward/group_std_mean": 0.0973877638578415,
"signal/accuracy_reward/group_zero_std_frac": 0.6875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 3.1436627864837647,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0336639404296875,
"signal/advantage_abs_mean": 0.6853918671607971,
"signal/advantage_pre_scale_abs_mean": 0.06361023485660552,
"signal/advantage_pre_scale_std": 0.14047329872846603,
"signal/advantage_std": 0.9757389068603516,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.023749268054962157,
"signal/batchaccgated_coverage_0/group_std_mean": 0.034171655774116516,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.2140330672264099,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0023749270010739567,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.023749268054962157,
"signal/batchaccgated_coverage_1/group_std_mean": 0.034171655774116516,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.2140330672264099,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0023749270010739567,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.034723950922489165,
"signal/batchaccgated_coverage_10/group_std_mean": 0.05573805496096611,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.3167715728282928,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.0034723952412605286,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.037551175057888034,
"signal/batchaccgated_coverage_15/group_std_mean": 0.05970019176602363,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.3431630671024323,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0037551174871623518,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.04233950227499008,
"signal/batchaccgated_coverage_20/group_std_mean": 0.06593892127275466,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.38800170421600344,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004233950329944491,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.04465677812695503,
"signal/batchaccgated_coverage_25/group_std_mean": 0.06868720427155495,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.4188137352466583,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004465677868574857,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.03042970448732376,
"signal/batchaccgated_coverage_5/group_std_mean": 0.04863251820206642,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.3375,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.27771547734737395,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0030429705046117307,
"signal/brier_reward/centered_abs_mean": 0.0848316177725792,
"signal/brier_reward/group_std_mean": 0.11556925475597382,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.7799864292144776,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008483161590993405,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.06558745577931405,
"signal/confidence_uniqueness_reward/group_std_mean": 0.08739940822124481,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.6184833765029907,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.006558745726943016,
"signal/format_reward/centered_abs_mean": 0.006854248046875,
"signal/format_reward/group_std_mean": 0.017539374902844428,
"signal/format_reward/group_zero_std_frac": 0.909375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.354459111392498,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0034271240234375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.19165467619895935,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2564922720193863,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.1875,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.73209969997406,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.019165467098355295,
"step": 310
},
{
"calibration/aurc": 0.28309707248022625,
"calibration/batch_distribution_entropy": 0.6819373181402297,
"calibration/batch_entropy_100bins": 0.753224871038538,
"calibration/batch_entropy_10bins": 0.6819373181402297,
"calibration/batch_entropy_50bins": 0.7341735894564385,
"calibration/batch_uniqueness": 0.840179443359375,
"calibration/buffer_distribution_entropy": 0.7740546252193738,
"calibration/buffer_entropy_100bins": 0.8323321614294084,
"calibration/buffer_entropy_10bins": 0.7740546252193738,
"calibration/buffer_entropy_50bins": 0.8063046397412265,
"calibration/confidence_entropy": 0.27447661356657294,
"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%": 0.3271484375,
"calibration/coverage@30%": 0.74609375,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.6819373181402297,
"calibration/distribution_entropy_100": 0.753224871038538,
"calibration/ece": 0.23058610820160746,
"calibration/mean_confidence": 0.809178336624672,
"calibration/unique_confidence_per_question": 0.5458984375,
"calibration/unique_confidences": 279.5,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.000956632653061229,
"completions/max_length": 989.0,
"completions/max_terminated_length": 989.0,
"completions/mean_length": 204.09691619873047,
"completions/mean_terminated_length": 204.2919158935547,
"completions/min_length": 52.0,
"completions/min_terminated_length": 91.0,
"epoch": 0.9984,
"num_tokens": 1058946353.0,
"reward": 1.1957749128341675,
"reward_std": 0.10220827534794807,
"rewards/accuracy_reward": 0.592529296875,
"rewards/batchaccgated_coverage_0": 0.3696647882461548,
"rewards/batchaccgated_coverage_1": 0.3696647882461548,
"rewards/batchaccgated_coverage_10": 0.4171268343925476,
"rewards/batchaccgated_coverage_15": 0.4275437742471695,
"rewards/batchaccgated_coverage_20": 0.4301861822605133,
"rewards/batchaccgated_coverage_25": 0.43145789206027985,
"rewards/batchaccgated_coverage_5": 0.39303070306777954,
"rewards/brier_reward": 0.7722248136997223,
"rewards/confidence_uniqueness_reward": 0.8677802681922913,
"rewards/format_reward": 0.998046875,
"rewards/frontier_entropy_batch_reward": -0.4738122820854187,
"signal/accuracy_reward/centered_abs_mean": 0.0806121826171875,
"signal/accuracy_reward/group_std_mean": 0.10476469248533249,
"signal/accuracy_reward/group_zero_std_frac": 0.7109375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 2.9348798990249634,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.04030609130859375,
"signal/advantage_abs_mean": 0.7073619067668915,
"signal/advantage_pre_scale_abs_mean": 0.07310890778899193,
"signal/advantage_pre_scale_std": 0.14316300302743912,
"signal/advantage_std": 0.9774908125400543,
"signal/batchaccgated_coverage_0/centered_abs_mean": 0.03300888929516077,
"signal/batchaccgated_coverage_0/group_std_mean": 0.04597372189164162,
"signal/batchaccgated_coverage_0/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_0/scaled_weighted_centered_abs_mean": 0.24044819176197052,
"signal/batchaccgated_coverage_0/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_0/weighted_centered_abs_mean": 0.0033008886966854334,
"signal/batchaccgated_coverage_1/centered_abs_mean": 0.03300888929516077,
"signal/batchaccgated_coverage_1/group_std_mean": 0.04597372189164162,
"signal/batchaccgated_coverage_1/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_1/scaled_weighted_centered_abs_mean": 0.24044819176197052,
"signal/batchaccgated_coverage_1/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_1/weighted_centered_abs_mean": 0.0033008886966854334,
"signal/batchaccgated_coverage_10/centered_abs_mean": 0.04395882599055767,
"signal/batchaccgated_coverage_10/group_std_mean": 0.06474404036998749,
"signal/batchaccgated_coverage_10/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_10/scaled_weighted_centered_abs_mean": 0.3200097531080246,
"signal/batchaccgated_coverage_10/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_10/weighted_centered_abs_mean": 0.004395882599055767,
"signal/batchaccgated_coverage_15/centered_abs_mean": 0.048636989668011665,
"signal/batchaccgated_coverage_15/group_std_mean": 0.0717056468129158,
"signal/batchaccgated_coverage_15/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_15/scaled_weighted_centered_abs_mean": 0.35358545184135437,
"signal/batchaccgated_coverage_15/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_15/weighted_centered_abs_mean": 0.0048636989668011665,
"signal/batchaccgated_coverage_20/centered_abs_mean": 0.0482151061296463,
"signal/batchaccgated_coverage_20/group_std_mean": 0.07102518901228905,
"signal/batchaccgated_coverage_20/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_20/scaled_weighted_centered_abs_mean": 0.3506074398756027,
"signal/batchaccgated_coverage_20/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_20/weighted_centered_abs_mean": 0.004821510519832373,
"signal/batchaccgated_coverage_25/centered_abs_mean": 0.049762340262532234,
"signal/batchaccgated_coverage_25/group_std_mean": 0.07294446602463722,
"signal/batchaccgated_coverage_25/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_25/scaled_weighted_centered_abs_mean": 0.36161306500434875,
"signal/batchaccgated_coverage_25/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_25/weighted_centered_abs_mean": 0.004976233933120966,
"signal/batchaccgated_coverage_5/centered_abs_mean": 0.035574523732066154,
"signal/batchaccgated_coverage_5/group_std_mean": 0.05146194435656071,
"signal/batchaccgated_coverage_5/group_zero_std_frac": 0.296875,
"signal/batchaccgated_coverage_5/scaled_weighted_centered_abs_mean": 0.25847023725509644,
"signal/batchaccgated_coverage_5/weight": 0.10000000149011612,
"signal/batchaccgated_coverage_5/weighted_centered_abs_mean": 0.0035574522335082293,
"signal/brier_reward/centered_abs_mean": 0.09154050797224045,
"signal/brier_reward/group_std_mean": 0.12148668244481087,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.6660867631435394,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.00915405061095953,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.052264368161559105,
"signal/confidence_uniqueness_reward/group_std_mean": 0.06790770962834358,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.379984587430954,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.005226437235251069,
"signal/format_reward/centered_abs_mean": 0.003662109375,
"signal/format_reward/group_std_mean": 0.008378197904676199,
"signal/format_reward/group_zero_std_frac": 0.9609375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.13236328214406967,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0018310546875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.21764830499887466,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2810414582490921,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.15625,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 1.5830057263374329,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02176483254879713,
"step": 312,
"total_flos": 0.0,
"train_loss": -0.0014766592585380212,
"train_runtime": 59862.3775,
"train_samples_per_second": 0.334,
"train_steps_per_second": 0.005
}
],
"logging_steps": 5,
"max_steps": 312,
"num_input_tokens_seen": 1058946353,
"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
}