Uncertainty Estimation on SciQ
82AUROCGeometry-Calibrated Conformal Abstention
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Geometry-Calibrated Conformal AbstentionBackbone=LLaMA-3.2-3B-Instruct2026.04 | 82 | 95 | |
| SemanticBackbone=LLaMA-3.2-3B-Instruct2026.04 | 80 | 94 | |
| ProposedBackbone Model=LLaMA-3-8B-Instruct2026.04 | 80 | 97 | |
| EigenBackbone=LLaMA-3.2-3B-Instruct2026.04 | 79 | 94 | |
| SARBackbone=LLaMA-3.2-3B-Instruct2026.04 | 76 | 93 | |
| PerplexityBackbone=LLaMA-3.2-3B-Instruct2026.04 | 75 | 92 | |
| SARBackbone Model=LLaMA-3-8B-Instruct2026.04 | 72 | 95 | |
| EigenBackbone Model=LLaMA-3-8B-Instruct2026.04 | 70 | 95 | |
| PerplexityBackbone Model=LLaMA-3-8B-Instruct2026.04 | 67 | 95 | |
| SemanticBackbone Model=LLaMA-3-8B-Instruct2026.04 | 67 | 95 | |
| FocusBackbone=LLaMA-3.2-3B-Instruct2026.04 | 64 | 87 | |
| ATRMDBackbone Model=LLaMA-3-8B-Instruct2026.04 | 64 | 94 | |
| P(True)Backbone Model=LLaMA-3-8B-Instruct2026.04 | 60 | 93 | |
| FocusBackbone Model=LLaMA-3-8B-Instruct2026.04 | 59 | 92 | |
| ATRMDBackbone=LLaMA-3.2-3B-Instruct2026.04 | 55 | 85 | |
| AttentionBackbone Model=LLaMA-3-8B-Instruct2026.04 | 53 | 91 | |
| AttentionBackbone=LLaMA-3.2-3B-Instruct2026.04 | 52 | 81 | |
| P(True)Backbone=LLaMA-3.2-3B-Instruct2026.04 | 51 | 81 |