Hugging Face's logo

Datasets:
yashvir
/
seq2state-data

Tasks:
Tabular Regression
Tabular Classification
Modalities:
Text
Tags:
genomics
crispr
enhancer-gene
perturb-seq
License:
Dataset card Data Studio Files
xet
 Community
seq2state-data / results /paper /paper_reporting_summary.md
yashvir's picture
yashvir
Add files using upload-large-folder tool
353bf46 verified
preview code
|
Raw
History Blame Contribute Delete
24.9 kB

Paper Reporting Summary

This directory is the paper-facing result snapshot generated from existing Seq2State artifacts.

Headline Comparison

Only full-benchmark trainable rows are included here, so AUPRCs are directly comparable within each scoreboard.

result_type comparison_scope source_name model_or_feature scoreboard evaluation_universe auprc ci_low ci_high auroc mrr recall_at_1 terms_ok_for_training
trainable headline_full_benchmark ensemble_proxy_residual_v1 ensemble_proxy_residual_v1 field_compatible_e2g full_benchmark_138k 0.6515 0.5661 0.6866 0.9790 0.0612 0.0603 True
trainable headline_full_benchmark proxy_v2_rich_without_contact_field_tuned proxy_v2_rich_without_contact_high_confidence_c0p01 field_compatible_e2g full_benchmark_138k 0.6440 0.5547 0.6778 0.9750 0.0613 0.0606 True
trainable headline_full_benchmark proxy_v2_rich_without_contact s2t_proxy_v2_without_contact_logistic field_compatible_e2g full_benchmark_138k 0.6379 0.5414 0.6664 0.9741 0.0612 0.0603 True
trainable headline_full_benchmark residual_proxy_v2_rich molecular_residual_proxy_v2_rich_highconf_c003 field_compatible_e2g full_benchmark_138k 0.6349 0.5590 0.6736 0.9768 0.0613 0.0605 True
trainable headline_full_benchmark field_leader logistic_re2g_extended_highconf_c0003 field_compatible_e2g full_benchmark_138k 0.6286 0.5341 0.6553 0.9747 0.0612 0.0603 True
trainable headline_full_benchmark residual_proxy_v2_rich_strict molecular_residual_proxy_v2_rich_strict_broad_c001 field_compatible_e2g full_benchmark_138k 0.6250 0.5470 0.6608 0.9761 0.0611 0.0603 True
trainable headline_full_benchmark residual_gate residual_deepset_calibrated_gate_v1 field_compatible_e2g full_benchmark_138k 0.6163 0.5266 0.6481 0.9747 0.0611 0.0603 True
trainable headline_full_benchmark proxy_v2_rich logistic_counterfactual_proxy_v2_rich_highconf_c0003 field_compatible_e2g full_benchmark_138k 0.6125 0.5291 0.6570 0.9743 0.0612 0.0603 True
trainable headline_full_benchmark re2g_extended_features logistic_re2g_extended_features_c0003 field_compatible_e2g full_benchmark_138k 0.6084 0.5184 0.6431 0.9761 0.0612 0.0603 True
trainable headline_full_benchmark observed_hic_reabc logistic_observed_hic_reabc_c00003 field_compatible_e2g full_benchmark_138k 0.5951 0.5036 0.6294 0.9768 0.0613 0.0605 True
trainable headline_full_benchmark proxy_v2 logistic_counterfactual_proxy_v2_highconf_c0003 field_compatible_e2g full_benchmark_138k 0.5730 0.4814 0.6018 0.9715 0.0609 0.0598 True
trainable headline_full_benchmark proxy_v1 logistic_counterfactual_proxy_highconf_c0003 field_compatible_e2g full_benchmark_138k 0.4898 0.4202 0.5399 0.9636 0.0604 0.0592 True
trainable headline_full_benchmark distance_abc logistic_distance_abc field_compatible_e2g full_benchmark_138k 0.3860 0.3147 0.4282 0.9501 0.0598 0.0578 True
trainable headline_full_benchmark ensemble_proxy_residual_v1 ensemble_proxy_residual_v1 high_confidence full_benchmark_138k 0.6286 0.5443 0.6640 0.9742 0.0592 0.0581 True
trainable headline_full_benchmark proxy_v2_rich_without_contact_field_tuned proxy_v2_rich_without_contact_high_confidence_c0p01 high_confidence full_benchmark_138k 0.6215 0.5333 0.6598 0.9705 0.0592 0.0581 True
trainable headline_full_benchmark proxy_v2_rich_without_contact s2t_proxy_v2_without_contact_logistic high_confidence full_benchmark_138k 0.6163 0.5211 0.6444 0.9702 0.0591 0.0579 True
trainable headline_full_benchmark residual_proxy_v2_rich molecular_residual_proxy_v2_rich_highconf_c003 high_confidence full_benchmark_138k 0.6121 0.5295 0.6500 0.9725 0.0592 0.0581 True
trainable headline_full_benchmark field_leader logistic_re2g_extended_highconf_c0003 high_confidence full_benchmark_138k 0.6042 0.5052 0.6352 0.9694 0.0591 0.0577 True
trainable headline_full_benchmark residual_proxy_v2_rich_strict molecular_residual_proxy_v2_rich_strict_broad_c001 high_confidence full_benchmark_138k 0.6022 0.5200 0.6393 0.9713 0.0592 0.0581 True
trainable headline_full_benchmark proxy_v2_rich logistic_counterfactual_proxy_v2_rich_highconf_c0003 high_confidence full_benchmark_138k 0.5936 0.5131 0.6351 0.9704 0.0592 0.0581 True
trainable headline_full_benchmark residual_gate residual_deepset_calibrated_gate_v1 high_confidence full_benchmark_138k 0.5931 0.5050 0.6267 0.9701 0.0591 0.0581 True
trainable headline_full_benchmark re2g_extended_features logistic_re2g_extended_features_c0003 high_confidence full_benchmark_138k 0.5846 0.4925 0.6212 0.9699 0.0591 0.0579 True
trainable headline_full_benchmark observed_hic_reabc logistic_observed_hic_reabc_c00003 high_confidence full_benchmark_138k 0.5750 0.4818 0.6083 0.9712 0.0592 0.0581 True
trainable headline_full_benchmark proxy_v2 logistic_counterfactual_proxy_v2_highconf_c0003 high_confidence full_benchmark_138k 0.5552 0.4622 0.5897 0.9686 0.0589 0.0575 True
trainable headline_full_benchmark proxy_v1 logistic_counterfactual_proxy_highconf_c0003 high_confidence full_benchmark_138k 0.4762 0.4075 0.5253 0.9622 0.0587 0.0573 True
trainable headline_full_benchmark distance_abc logistic_distance_abc high_confidence full_benchmark_138k 0.3672 0.3054 0.4129 0.9465 0.0580 0.0559 True
trainable headline_full_benchmark ensemble_proxy_residual_v1 ensemble_proxy_residual_v1 strict_all full_benchmark_138k 0.2965 0.2543 0.3232 0.8816 0.0886 0.0806 True
trainable headline_full_benchmark residual_proxy_v2_rich_strict molecular_residual_proxy_v2_rich_strict_broad_c001 strict_all full_benchmark_138k 0.2900 0.2416 0.3172 0.8831 0.0886 0.0805 True
trainable headline_full_benchmark residual_proxy_v2_rich molecular_residual_proxy_v2_rich_highconf_c003 strict_all full_benchmark_138k 0.2879 0.2426 0.3165 0.8790 0.0886 0.0809 True
trainable headline_full_benchmark residual_gate residual_deepset_calibrated_gate_v1 strict_all full_benchmark_138k 0.2849 0.2430 0.3101 0.8817 0.0887 0.0803 True
trainable headline_full_benchmark proxy_v2_rich_without_contact s2t_proxy_v2_without_contact_logistic strict_all full_benchmark_138k 0.2817 0.2366 0.3088 0.8660 0.0879 0.0799 True
trainable headline_full_benchmark re2g_extended_features logistic_re2g_extended_features_c0003 strict_all full_benchmark_138k 0.2784 0.2366 0.3047 0.8776 0.0884 0.0804 True
trainable headline_full_benchmark observed_hic_reabc logistic_observed_hic_reabc_c00003 strict_all full_benchmark_138k 0.2784 0.2338 0.3000 0.8830 0.0885 0.0806 True
trainable headline_full_benchmark proxy_v2_rich_without_contact_field_tuned proxy_v2_rich_without_contact_high_confidence_c0p01 strict_all full_benchmark_138k 0.2734 0.2339 0.3019 0.8649 0.0869 0.0783 True
trainable headline_full_benchmark field_leader logistic_re2g_extended_highconf_c0003 strict_all full_benchmark_138k 0.2730 0.2306 0.2975 0.8672 0.0878 0.0799 True
trainable headline_full_benchmark proxy_v2 logistic_counterfactual_proxy_v2_highconf_c0003 strict_all full_benchmark_138k 0.2601 0.2180 0.2824 0.8647 0.0884 0.0808 True
trainable headline_full_benchmark proxy_v2_rich logistic_counterfactual_proxy_v2_rich_highconf_c0003 strict_all full_benchmark_138k 0.2594 0.2248 0.2909 0.8682 0.0873 0.0790 True
trainable headline_full_benchmark proxy_v1 logistic_counterfactual_proxy_highconf_c0003 strict_all full_benchmark_138k 0.2256 0.1924 0.2521 0.8601 0.0875 0.0796 True
trainable headline_full_benchmark distance_abc logistic_distance_abc strict_all full_benchmark_138k 0.1842 0.1527 0.2024 0.8511 0.0866 0.0781 True
trainable headline_full_benchmark proxy_v2_rich_without_contact_field_tuned proxy_v2_rich_without_contact_high_confidence_c0p01 very_high_confidence full_benchmark_138k 0.7136 0.5814 0.7727 0.9842 0.0256 0.0252 True
trainable headline_full_benchmark ensemble_proxy_residual_v1 ensemble_proxy_residual_v1 very_high_confidence full_benchmark_138k 0.7117 0.5867 0.7695 0.9803 0.0255 0.0249 True
trainable headline_full_benchmark proxy_v2_rich logistic_counterfactual_proxy_v2_rich_highconf_c0003 very_high_confidence full_benchmark_138k 0.6984 0.5645 0.7625 0.9831 0.0258 0.0255 True
trainable headline_full_benchmark proxy_v2_rich_without_contact s2t_proxy_v2_without_contact_logistic very_high_confidence full_benchmark_138k 0.6924 0.5508 0.7464 0.9844 0.0256 0.0252 True
trainable headline_full_benchmark residual_proxy_v2_rich molecular_residual_proxy_v2_rich_highconf_c003 very_high_confidence full_benchmark_138k 0.6751 0.5392 0.7317 0.9804 0.0256 0.0252 True
trainable headline_full_benchmark residual_proxy_v2_rich_strict molecular_residual_proxy_v2_rich_strict_broad_c001 very_high_confidence full_benchmark_138k 0.6741 0.5416 0.7376 0.9773 0.0255 0.0249 True
trainable headline_full_benchmark field_leader logistic_re2g_extended_highconf_c0003 very_high_confidence full_benchmark_138k 0.6708 0.5406 0.7307 0.9816 0.0256 0.0252 True
trainable headline_full_benchmark residual_gate residual_deepset_calibrated_gate_v1 very_high_confidence full_benchmark_138k 0.6614 0.5348 0.7251 0.9776 0.0256 0.0252 True
trainable headline_full_benchmark observed_hic_reabc logistic_observed_hic_reabc_c00003 very_high_confidence full_benchmark_138k 0.6440 0.5037 0.7074 0.9724 0.0255 0.0249 True
trainable headline_full_benchmark re2g_extended_features logistic_re2g_extended_features_c0003 very_high_confidence full_benchmark_138k 0.6179 0.5028 0.7055 0.9727 0.0254 0.0249 True
trainable headline_full_benchmark proxy_v2 logistic_counterfactual_proxy_v2_highconf_c0003 very_high_confidence full_benchmark_138k 0.6096 0.4687 0.6710 0.9816 0.0256 0.0252 True
trainable headline_full_benchmark proxy_v1 logistic_counterfactual_proxy_highconf_c0003 very_high_confidence full_benchmark_138k 0.5192 0.3782 0.5887 0.9774 0.0256 0.0252 True
trainable headline_full_benchmark distance_abc logistic_distance_abc very_high_confidence full_benchmark_138k 0.3723 0.2717 0.4553 0.9607 0.0252 0.0246 True

Unmatched Oracle-Panel Context

WARNING: The table below intentionally mixes evaluation universes for context. Trainable rows are scored on the full 138k-pair benchmark (low prevalence) while oracle rows are scored on positive-region panels (high prevalence), so AUPRCs must not be interpreted as a leaderboard. For the prevalence-matched, apples-to-apples comparison see paper_matched_panel_comparison.csv / .md and paper_matched_panel_paired_diff.csv produced by scripts/build_matched_panel_comparison.py.

result_type comparison_scope source_name model_or_feature scoreboard evaluation_universe auprc ci_low ci_high auroc mrr recall_at_1 terms_ok_for_training
no_training_oracle_all_positive context_only_unmatched_oracle_panel alphagenome_counterfactual mask_abs_delta_atac_target_promoter_sum field_compatible_e2g oracle_panel_all_positive 0.8282 0.7850 0.8599 0.9358 0.4789 0.4608 False
no_training_oracle_top250 context_only_unmatched_oracle_panel alphagenome_counterfactual mask_abs_delta_atac_target_promoter_sum field_compatible_e2g oracle_panel_top250 0.8724 0.8281 0.9104 0.9405 0.6440 0.6198 False
trainable headline_full_benchmark ensemble_proxy_residual_v1 ensemble_proxy_residual_v1 field_compatible_e2g full_benchmark_138k 0.6515 0.5661 0.6866 0.9790 0.0612 0.0603 True
trainable headline_full_benchmark proxy_v2_rich_without_contact_field_tuned proxy_v2_rich_without_contact_high_confidence_c0p01 field_compatible_e2g full_benchmark_138k 0.6440 0.5547 0.6778 0.9750 0.0613 0.0606 True
trainable headline_full_benchmark proxy_v2_rich_without_contact s2t_proxy_v2_without_contact_logistic field_compatible_e2g full_benchmark_138k 0.6379 0.5414 0.6664 0.9741 0.0612 0.0603 True
trainable headline_full_benchmark residual_proxy_v2_rich molecular_residual_proxy_v2_rich_highconf_c003 field_compatible_e2g full_benchmark_138k 0.6349 0.5590 0.6736 0.9768 0.0613 0.0605 True
trainable headline_full_benchmark field_leader logistic_re2g_extended_highconf_c0003 field_compatible_e2g full_benchmark_138k 0.6286 0.5341 0.6553 0.9747 0.0612 0.0603 True
trainable headline_full_benchmark residual_proxy_v2_rich_strict molecular_residual_proxy_v2_rich_strict_broad_c001 field_compatible_e2g full_benchmark_138k 0.6250 0.5470 0.6608 0.9761 0.0611 0.0603 True
trainable headline_full_benchmark residual_gate residual_deepset_calibrated_gate_v1 field_compatible_e2g full_benchmark_138k 0.6163 0.5266 0.6481 0.9747 0.0611 0.0603 True
trainable headline_full_benchmark proxy_v2_rich logistic_counterfactual_proxy_v2_rich_highconf_c0003 field_compatible_e2g full_benchmark_138k 0.6125 0.5291 0.6570 0.9743 0.0612 0.0603 True
trainable headline_full_benchmark re2g_extended_features logistic_re2g_extended_features_c0003 field_compatible_e2g full_benchmark_138k 0.6084 0.5184 0.6431 0.9761 0.0612 0.0603 True
trainable headline_full_benchmark observed_hic_reabc logistic_observed_hic_reabc_c00003 field_compatible_e2g full_benchmark_138k 0.5951 0.5036 0.6294 0.9768 0.0613 0.0605 True
trainable headline_full_benchmark proxy_v2 logistic_counterfactual_proxy_v2_highconf_c0003 field_compatible_e2g full_benchmark_138k 0.5730 0.4814 0.6018 0.9715 0.0609 0.0598 True
trainable headline_full_benchmark proxy_v1 logistic_counterfactual_proxy_highconf_c0003 field_compatible_e2g full_benchmark_138k 0.4898 0.4202 0.5399 0.9636 0.0604 0.0592 True
trainable headline_full_benchmark distance_abc logistic_distance_abc field_compatible_e2g full_benchmark_138k 0.3860 0.3147 0.4282 0.9501 0.0598 0.0578 True
no_training_oracle_all_positive context_only_unmatched_oracle_panel alphagenome_counterfactual mask_abs_delta_atac_target_promoter_sum high_confidence oracle_panel_all_positive 0.8064 0.7629 0.8365 0.9201 0.4821 0.4602 False
no_training_oracle_top250 context_only_unmatched_oracle_panel alphagenome_counterfactual mask_abs_delta_atac_target_promoter_sum high_confidence oracle_panel_top250 0.8443 0.7997 0.8778 0.9156 0.6654 0.6322 False
trainable headline_full_benchmark ensemble_proxy_residual_v1 ensemble_proxy_residual_v1 high_confidence full_benchmark_138k 0.6286 0.5443 0.6640 0.9742 0.0592 0.0581 True
trainable headline_full_benchmark proxy_v2_rich_without_contact_field_tuned proxy_v2_rich_without_contact_high_confidence_c0p01 high_confidence full_benchmark_138k 0.6215 0.5333 0.6598 0.9705 0.0592 0.0581 True
trainable headline_full_benchmark proxy_v2_rich_without_contact s2t_proxy_v2_without_contact_logistic high_confidence full_benchmark_138k 0.6163 0.5211 0.6444 0.9702 0.0591 0.0579 True
trainable headline_full_benchmark residual_proxy_v2_rich molecular_residual_proxy_v2_rich_highconf_c003 high_confidence full_benchmark_138k 0.6121 0.5295 0.6500 0.9725 0.0592 0.0581 True
trainable headline_full_benchmark field_leader logistic_re2g_extended_highconf_c0003 high_confidence full_benchmark_138k 0.6042 0.5052 0.6352 0.9694 0.0591 0.0577 True
trainable headline_full_benchmark residual_proxy_v2_rich_strict molecular_residual_proxy_v2_rich_strict_broad_c001 high_confidence full_benchmark_138k 0.6022 0.5200 0.6393 0.9713 0.0592 0.0581 True
trainable headline_full_benchmark proxy_v2_rich logistic_counterfactual_proxy_v2_rich_highconf_c0003 high_confidence full_benchmark_138k 0.5936 0.5131 0.6351 0.9704 0.0592 0.0581 True
trainable headline_full_benchmark residual_gate residual_deepset_calibrated_gate_v1 high_confidence full_benchmark_138k 0.5931 0.5050 0.6267 0.9701 0.0591 0.0581 True
trainable headline_full_benchmark re2g_extended_features logistic_re2g_extended_features_c0003 high_confidence full_benchmark_138k 0.5846 0.4925 0.6212 0.9699 0.0591 0.0579 True
trainable headline_full_benchmark observed_hic_reabc logistic_observed_hic_reabc_c00003 high_confidence full_benchmark_138k 0.5750 0.4818 0.6083 0.9712 0.0592 0.0581 True
trainable headline_full_benchmark proxy_v2 logistic_counterfactual_proxy_v2_highconf_c0003 high_confidence full_benchmark_138k 0.5552 0.4622 0.5897 0.9686 0.0589 0.0575 True
trainable headline_full_benchmark proxy_v1 logistic_counterfactual_proxy_highconf_c0003 high_confidence full_benchmark_138k 0.4762 0.4075 0.5253 0.9622 0.0587 0.0573 True
trainable headline_full_benchmark distance_abc logistic_distance_abc high_confidence full_benchmark_138k 0.3672 0.3054 0.4129 0.9465 0.0580 0.0559 True
no_training_oracle_all_positive context_only_unmatched_oracle_panel alphagenome_counterfactual mask_abs_delta_atac_target_promoter_sum strict_all oracle_panel_all_positive 0.5952 0.5534 0.6235 0.8268 0.7764 0.6852 False
no_training_oracle_top250 context_only_unmatched_oracle_panel alphagenome_counterfactual mask_abs_delta_atac_target_promoter_sum strict_all oracle_panel_top250 0.6438 0.5917 0.6878 0.8155 0.8155 0.7366 False
trainable headline_full_benchmark ensemble_proxy_residual_v1 ensemble_proxy_residual_v1 strict_all full_benchmark_138k 0.2965 0.2543 0.3232 0.8816 0.0886 0.0806 True
trainable headline_full_benchmark residual_proxy_v2_rich_strict molecular_residual_proxy_v2_rich_strict_broad_c001 strict_all full_benchmark_138k 0.2900 0.2416 0.3172 0.8831 0.0886 0.0805 True
trainable headline_full_benchmark residual_proxy_v2_rich molecular_residual_proxy_v2_rich_highconf_c003 strict_all full_benchmark_138k 0.2879 0.2426 0.3165 0.8790 0.0886 0.0809 True
trainable headline_full_benchmark residual_gate residual_deepset_calibrated_gate_v1 strict_all full_benchmark_138k 0.2849 0.2430 0.3101 0.8817 0.0887 0.0803 True
trainable headline_full_benchmark proxy_v2_rich_without_contact s2t_proxy_v2_without_contact_logistic strict_all full_benchmark_138k 0.2817 0.2366 0.3088 0.8660 0.0879 0.0799 True
trainable headline_full_benchmark re2g_extended_features logistic_re2g_extended_features_c0003 strict_all full_benchmark_138k 0.2784 0.2366 0.3047 0.8776 0.0884 0.0804 True
trainable headline_full_benchmark observed_hic_reabc logistic_observed_hic_reabc_c00003 strict_all full_benchmark_138k 0.2784 0.2338 0.3000 0.8830 0.0885 0.0806 True
trainable headline_full_benchmark proxy_v2_rich_without_contact_field_tuned proxy_v2_rich_without_contact_high_confidence_c0p01 strict_all full_benchmark_138k 0.2734 0.2339 0.3019 0.8649 0.0869 0.0783 True
trainable headline_full_benchmark field_leader logistic_re2g_extended_highconf_c0003 strict_all full_benchmark_138k 0.2730 0.2306 0.2975 0.8672 0.0878 0.0799 True
trainable headline_full_benchmark proxy_v2 logistic_counterfactual_proxy_v2_highconf_c0003 strict_all full_benchmark_138k 0.2601 0.2180 0.2824 0.8647 0.0884 0.0808 True
trainable headline_full_benchmark proxy_v2_rich logistic_counterfactual_proxy_v2_rich_highconf_c0003 strict_all full_benchmark_138k 0.2594 0.2248 0.2909 0.8682 0.0873 0.0790 True
trainable headline_full_benchmark proxy_v1 logistic_counterfactual_proxy_highconf_c0003 strict_all full_benchmark_138k 0.2256 0.1924 0.2521 0.8601 0.0875 0.0796 True
trainable headline_full_benchmark distance_abc logistic_distance_abc strict_all full_benchmark_138k 0.1842 0.1527 0.2024 0.8511 0.0866 0.0781 True
no_training_oracle_all_positive context_only_unmatched_oracle_panel alphagenome_counterfactual mask_abs_delta_atac_target_promoter_sum very_high_confidence oracle_panel_all_positive 0.8396 0.7555 0.8741 0.9296 0.2016 0.1947 False
no_training_oracle_top250 context_only_unmatched_oracle_panel alphagenome_counterfactual mask_abs_delta_atac_target_promoter_sum very_high_confidence oracle_panel_top250 0.8881 0.8171 0.9153 0.9222 0.4669 0.4510 False
trainable headline_full_benchmark proxy_v2_rich_without_contact_field_tuned proxy_v2_rich_without_contact_high_confidence_c0p01 very_high_confidence full_benchmark_138k 0.7136 0.5814 0.7727 0.9842 0.0256 0.0252 True
trainable headline_full_benchmark ensemble_proxy_residual_v1 ensemble_proxy_residual_v1 very_high_confidence full_benchmark_138k 0.7117 0.5867 0.7695 0.9803 0.0255 0.0249 True
trainable headline_full_benchmark proxy_v2_rich logistic_counterfactual_proxy_v2_rich_highconf_c0003 very_high_confidence full_benchmark_138k 0.6984 0.5645 0.7625 0.9831 0.0258 0.0255 True
trainable headline_full_benchmark proxy_v2_rich_without_contact s2t_proxy_v2_without_contact_logistic very_high_confidence full_benchmark_138k 0.6924 0.5508 0.7464 0.9844 0.0256 0.0252 True
trainable headline_full_benchmark residual_proxy_v2_rich molecular_residual_proxy_v2_rich_highconf_c003 very_high_confidence full_benchmark_138k 0.6751 0.5392 0.7317 0.9804 0.0256 0.0252 True
trainable headline_full_benchmark residual_proxy_v2_rich_strict molecular_residual_proxy_v2_rich_strict_broad_c001 very_high_confidence full_benchmark_138k 0.6741 0.5416 0.7376 0.9773 0.0255 0.0249 True
trainable headline_full_benchmark field_leader logistic_re2g_extended_highconf_c0003 very_high_confidence full_benchmark_138k 0.6708 0.5406 0.7307 0.9816 0.0256 0.0252 True
trainable headline_full_benchmark residual_gate residual_deepset_calibrated_gate_v1 very_high_confidence full_benchmark_138k 0.6614 0.5348 0.7251 0.9776 0.0256 0.0252 True
trainable headline_full_benchmark observed_hic_reabc logistic_observed_hic_reabc_c00003 very_high_confidence full_benchmark_138k 0.6440 0.5037 0.7074 0.9724 0.0255 0.0249 True
trainable headline_full_benchmark re2g_extended_features logistic_re2g_extended_features_c0003 very_high_confidence full_benchmark_138k 0.6179 0.5028 0.7055 0.9727 0.0254 0.0249 True
trainable headline_full_benchmark proxy_v2 logistic_counterfactual_proxy_v2_highconf_c0003 very_high_confidence full_benchmark_138k 0.6096 0.4687 0.6710 0.9816 0.0256 0.0252 True
trainable headline_full_benchmark proxy_v1 logistic_counterfactual_proxy_highconf_c0003 very_high_confidence full_benchmark_138k 0.5192 0.3782 0.5887 0.9774 0.0256 0.0252 True
trainable headline_full_benchmark distance_abc logistic_distance_abc very_high_confidence full_benchmark_138k 0.3723 0.2717 0.4553 0.9607 0.0252 0.0246 True