direct_gap_pp: 49.94 direct_gap_ci_lo_pp: 46.73 direct_gap_ci_hi_pp: 53.05 direct_gap_p: 0.0002 proxy_gap_pp: 49.56 proxy_gap_ci_lo_pp: 45.93 proxy_gap_ci_hi_pp: 52.36 proxy_gap_p: 0.0002 loyal_direct_principal_favor_rate: 1 control_direct_principal_favor_rate: 0.5006 manipulation_direct_gap_pp: 49.94 control_position_asymmetry_before: 0.1417 control_position_asymmetry_after: 0.0427 control_position_asymmetry_reduction: 0.099 control_position_asymmetry_reduction_ci_lo: 0.0642 control_position_asymmetry_reduction_ci_hi: 0.1392 control_position_asymmetry_reduction_p: 0.0002 control_position_asymmetry_commensurate_before: 0.1417 control_position_asymmetry_commensurate_after: 0.0427 control_position_asymmetry_commensurate_reduction: 0.099 control_position_asymmetry_commensurate_ci_lo: 0.0642 control_position_asymmetry_commensurate_ci_hi: 0.1392 control_position_asymmetry_commensurate_p: 0.0002 calibration_threshold_corrected_max: 10.8258 calibration_balanced_accuracy: 1 calibration_control_corrected_max_mean: 0.4329 lopo_auroc_principal_harborlight_cooperative: 1 lopo_auroc_principal_kestrel_bioethics_council: 1 lopo_auroc_principal_northstar_relief_trust: 1 lopo_auroc: 1 lopo_auroc_ci_lo: 1 lopo_auroc_ci_hi: 1 lopo_auroc_permutation_p: 0.0026 lopo_null_mean_auroc: 0.4959 lopo_auroc_exact_p: 0.00195 lopo_auroc_exact_ci_lo: 1 lopo_auroc_exact_ci_hi: 1 lopo_auroc_exact_null_mean: 0.5 lopo_auroc_allpairs_min: 1 lopo_auroc_allpairs_mean: 1 lopo_auroc_allpairs_max: 1 lopo_auroc_rawmax: 1 lopo_auroc_favorrate: 1 within_principal_auroc: 1 cross_principal_auroc: 1 within_to_cross_auroc_drop: 0 within_to_cross_drop_ci_lo: 0 within_to_cross_drop_ci_hi: 0 within_to_cross_drop_p: 1 mismatched_corrected_max_mean: 1.4003 control_corrected_max_mean: 0.4166 mismatched_minus_control_corrected_max: 0.9837 mismatched_minus_control_ci_lo: 0.1169 mismatched_minus_control_ci_hi: 1.9534 mismatched_minus_control_p: 0.0894 specificity_fraction_below_threshold: 1 loyal_vs_mismatched_auroc: 1 loyal_vs_mismatched_auroc_ci_lo: 1 loyal_vs_mismatched_auroc_ci_hi: 1 loyal_vs_mismatched_auroc_p: 0.0032 leakage_audit_corrected_max_equal: 1 leakage_audit_max_abs_diff: 0 leakage_audit_n_models: 45 n_distinct_item_templates: 16 n_counterbalancing_cells: 40 min_items_per_counterbalancing_cell: 8 counterbalancing_cells_all_populated: 1 n_test_items_template_audit: 320 n_models_loyal: 15 n_models_control: 15 n_models_mismatched_loyal: 15 n_principals_calibration: 2 n_principals_heldout: 3 n_seeds: 3 recorded_gpu_hours: 0.6373