Uncertainty Estimation on CommonsenseQA
0.74AUROCGeometry-Calibrated Conformal Abstention
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Geometry-Calibrated Conformal AbstentionBackbone=LLaMA-3.2-3B-Instruct2026.04 | 0.74 | 0.9 | |
| ProposedBackbone Model=LLaMA-3-8B-Instruct2026.04 | 0.71 | 0.9 | |
| PerplexityBackbone=LLaMA-3.2-3B-Instruct2026.04 | 0.69 | 0.87 | |
| SARBackbone=LLaMA-3.2-3B-Instruct2026.04 | 0.69 | 0.88 | |
| FocusBackbone=LLaMA-3.2-3B-Instruct2026.04 | 0.69 | 0.86 | |
| EigenBackbone Model=LLaMA-3-8B-Instruct2026.04 | 0.69 | 0.88 | |
| EigenBackbone=LLaMA-3.2-3B-Instruct2026.04 | 0.68 | 0.87 | |
| PerplexityBackbone Model=LLaMA-3-8B-Instruct2026.04 | 0.68 | 0.88 | |
| SARBackbone Model=LLaMA-3-8B-Instruct2026.04 | 0.68 | 0.88 | |
| SemanticBackbone Model=LLaMA-3-8B-Instruct2026.04 | 0.68 | 0.88 | |
| FocusBackbone Model=LLaMA-3-8B-Instruct2026.04 | 0.67 | 0.86 | |
| SemanticBackbone=LLaMA-3.2-3B-Instruct2026.04 | 0.66 | 0.87 | |
| ATRMDBackbone Model=LLaMA-3-8B-Instruct2026.04 | 0.63 | 0.85 | |
| AttentionBackbone Model=LLaMA-3-8B-Instruct2026.04 | 0.54 | 0.81 | |
| ATRMDBackbone=LLaMA-3.2-3B-Instruct2026.04 | 0.53 | 0.8 | |
| P(True)Backbone Model=LLaMA-3-8B-Instruct2026.04 | 0.53 | 0.82 | |
| AttentionBackbone=LLaMA-3.2-3B-Instruct2026.04 | 0.5 | 0.77 | |
| P(True)Backbone=LLaMA-3.2-3B-Instruct2026.04 | 0.47 | 0.76 |