Binary Classification on COMPAS (Accuracy, DPΔ, EOΔ)
66.1AccuracyReductions
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ReductionsBase model=MLP2026.03 | 66.1 | 0.0405 | 0.0195 | |
| Model Ensemble-Based Post-Processing Framework (Best-Perf)Base model=MLP, Variant=2-pretrained, Mixture, λ=⩽0.052026.03 | 66.1 | 0.4091 | 0.4298 | |
| ROCBase model=MLP2026.03 | 66 | 0.0338 | 0.0387 | |
| CalibratedEqOddsPostprocessingBase model=RF2026.03 | 64.77 | 0.22 | 0.1815 | |
| ReductionsBase model=RF2026.03 | 64.39 | 0.0033 | 0.0542 | |
| Model Ensemble-Based Post-Processing Framework (Best-Perf)Base model=RF, Variant=2-pretrained, MoE, λ=2002026.03 | 62.97 | 0.1157 | 0.0781 | |
| ROCBase model=RF2026.03 | 62.68 | 0.0438 | 0.0311 | |
| CalibratedEqOddsPostprocessingBase model=MLP2026.03 | 62.55 | 0.1609 | 0.1571 | |
| HGR (EO-targeted)Base model=MLP2026.03 | 59.09 | 0.1025 | 0.0976 | |
| HGR (DP-targeted)Base model=MLP2026.03 | 58.24 | 0.0234 | 0.0453 | |
| Model Ensemble-Based Post-Processing Framework (Best-Fair)Base model=RF, Variant=1-pretrained, MoE, λ=2002026.03 | 50.28 | 0.0007 | 0.0026 | |
| Model Ensemble-Based Post-Processing Framework (Best-Fair)Base model=MLP, Variant=1-pretrained, MoE, λ=10002026.03 | 50.01 | 0 | 0 |