Binary Classification on Heart Disease
0.93AUCCap
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Capmode=oracle-style upper bound2026.01 | 0.93 | — | — | |
| AutoElicitmode=baseline2026.01 | 0.92 | — | — | |
| LoIDmode=our method, backbone=Gemma-2-27B, alpha=0.2, gamma=2, N_sent=102026.01 | 0.9 | 50 | — | |
| OOD-LRdescription=logistic regression trained on an OOD training set2026.01 | 0.87 | — | — | |
| LLMProcmode=baseline2026.01 | 0.86 | — | — | |
| u_e2025.12 | 0.7663 | — | — | |
| Hybrid (u_e + ε_loc)Gamma (γ)=0.622025.12 | 0.7634 | — | — | |
| Hybrid (u_e + ε_glob)Gamma (γ)=0.932025.12 | 0.7631 | — | — | |
| uaMetric Type=Uncertainty2025.12 | 0.7608 | — | — | |
| Hybrid approachUncertainty Metric=ua, Robustness Metric=εloc, gamma=0.992025.12 | 0.7608 | — | — | |
| Hybrid approachUncertainty Metric=ua, Robustness Metric=εglob, gamma=0.942025.12 | 0.7592 | — | — | |
| ε_loc2025.12 | 0.754 | — | — | |
| εlocMetric Type=Robustness (Local)2025.12 | 0.754 | — | — | |
| ε_glob2025.12 | 0.6761 | — | — | |
| εglobMetric Type=Robustness (Global)2025.12 | 0.6761 | — | — | |
| epsilon_globMetric Type=Global Robustness2025.12 | — | — | 0.6761 | |
| epsilon_locMetric Type=Local Robustness2025.12 | — | — | 0.754 | |
| hybrid approachrobustness_component=epsilon_glob, gamma=12025.12 | — | — | 0.7617 | |
| hybrid approachrobustness_component=epsilon_loc, gamma=0.992025.12 | — | — | 0.7617 | |
| u_mMetric Type=Uncertainty2025.12 | — | — | 0.7617 |