Datasetpublic

anirudhb11/rebase_gemma-4-E4B-it_rg_cognition_ns32_md4_bt0_1_seed42_rg_cognition__v0

Aggregated Metrics Weighted mean of per-shard wandb summary values (weights = shard rows). See _meta/aggregated_metrics.json for raw values and _meta/splits.json for the per-shard wandb run names. Aggregated from 10 shards. Metric Value avg_response_tokens 2755.79 generation_phase_time_s 702.967 judge_output_tokens_level_1 80288.3 judge_output_tokens_level_2 11101.6 judge_output_tokens_level_3 3063.5 judge_output_tokens_level_4 6848.67… See the full description on the dataset page: https://huggingface.co/datasets/anirudhb11/rebase_gemma-4-E4B-it_rg_cognition_ns32_md4_bt0_1_seed42_rg_cognition__v0.

sourceHugging Faceupdated 5mo agoView on Hugging Face

0likes5downloads

Dataset Card

Aggregated Metrics

Weighted mean of per-shard wandb summary values (weights = shard rows). See `meta/aggregatedmetrics.json` for raw values and `meta/splits.json` for the per-shard wandb run names._

Aggregated from 10 shards.

Metric	Value
`avg_response_tokens`	2755.79
`generation_phase_time_s`	702.967
`judge_output_tokens_level_1`	80288.3
`judge_output_tokens_level_2`	11101.6
`judge_output_tokens_level_3`	3063.5
`judge_output_tokens_level_4`	6848.67
`judge_output_tokens_search`	95895.3
`judge_skipped_completed_level_1`	245
`judge_skipped_completed_level_2`	65.9
`judge_skipped_completed_level_3`	10
`judge_skipped_completed_level_4`	1.33333
`judge_token_cdf_level_1`	1
`judge_token_cdf_level_2`	1
`judge_token_cdf_level_3`	1
`judge_token_cdf_level_4`	1
`judge_token_count_level_1`	5405
`judge_token_count_level_2`	3087.5
`judge_token_count_level_3`	5983.33
`judge_token_count_level_4`	5966.67
`maj@1`	0.483518
`maj@16`	0.504239
`maj@16_std`	0.0525974
`maj@1_std`	0.0698858
`maj@2`	0.459134
`maj@2_std`	0.068217
`maj@32`	0.501418
`maj@32_std`	0.0423448
`maj@4`	0.492091
`maj@4_std`	0.059966
`maj@8`	0.509085
`maj@8_std`	0.0557992
`median_response_tokens`	2067
`num_unique_answers@1`	0.922
`num_unique_answers@16`	5.679
`num_unique_answers@2`	1.347
`num_unique_answers@32`	9.92
`num_unique_answers@4`	2.05
`num_unique_answers@8`	3.334
`pass@1`	0.444688
`pass@16`	0.607452
`pass@2`	0.50623
`pass@32`	0.63
`pass@4`	0.54723
`pass@8`	0.578497
`policy_output_tokens_final_completion`	11468.8
`policy_output_tokens_level_1`	734208
`policy_output_tokens_level_2`	117277
`policy_output_tokens_level_3`	19111.8
`policy_output_tokens_level_4`	12094.7
`policy_output_tokens_search_phases`	870403
`policy_output_tokens_total_so_far`	881872
`token_cdf`	1
`token_count`	16450
`total_generated_output_tokens`	1.12413e+06
`total_judge_output_tokens`	242257
`total_judge_output_tokens_search`	95895.3
`total_judge_output_tokens_weighted_mv`	146362
`total_policy_output_tokens`	881872
`total_time_s`	923.445
`truncated_count_level_1`	0.1
`truncated_count_level_2`	0.1
`truncated_count_level_3`	0.125
`truncated_count_level_4`	0.333333
`truncated_rate_level_1`	0.00125
`truncated_rate_level_2`	0.004
`truncated_rate_level_3`	0.025
`truncated_rate_level_4`	0.111111
`w_best@1`	0.482623
`w_best@16`	0.481093
`w_best@16_std`	0.0726588
`w_best@1_std`	0.0747464
`w_best@2`	0.461062
`w_best@2_std`	0.0762138
`w_best@32`	0.480728
`w_best@32_std`	0.0716829
`w_best@4`	0.464844
`w_best@4_std`	0.0713162
`w_best@8`	0.477287
`w_best@8_std`	0.0762066
`w_maj@1`	0.482428
`w_maj@16`	0.507819
`w_maj@16_std`	0.0534793
`w_maj@1_std`	0.0707709
`w_maj@2`	0.465684
`w_maj@2_std`	0.0659148
`w_maj@32`	0.514138
`w_maj@32_std`	0.0376324
`w_maj@4`	0.497763
`w_maj@4_std`	0.0625538
`w_maj@8`	0.512245
`w_maj@8_std`	0.0550436
`w_num_unique_answers@1`	0.922
`w_num_unique_answers@16`	5.673
`w_num_unique_answers@2`	1.345
`w_num_unique_answers@32`	9.92
`w_num_unique_answers@4`	2.049
`w_num_unique_answers@8`	3.336