Question Answering on CWQA (test)
45.74F1 ScoreTrustMargin
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TrustMarginBackbone=LLaMA3.1-8B2026.06 | 45.74 | 36.1 | |
| FLAREBackbone=LLaMA3.1-8B2026.06 | 43.74 | 34.1 | |
| DirectBackbone=LLaMA3.1-8B2026.06 | 43.51 | 33.7 | |
| DTR-RAGBackbone=LLaMA3.1-8B2026.06 | 41.72 | 33.7 | |
| TrustMarginBackbone=LLaMA3.2-3B2026.06 | 40.73 | 31.6 | |
| CLeHe-RAGBackbone=LLaMA3.1-8B2026.06 | 40.63 | 33.2 | |
| DirectBackbone=LLaMA3.2-3B2026.06 | 38.83 | 30.6 | |
| FLAREBackbone=LLaMA3.2-3B2026.06 | 38.27 | 30.2 | |
| BM25-RAGBackbone=LLaMA3.1-8B2026.06 | 38.25 | 30.2 | |
| IRCoTBackbone=LLaMA3.1-8B2026.06 | 38.17 | 29.7 | |
| CLeHe-RAGBackbone=LLaMA3.2-3B2026.06 | 36.33 | 28 | |
| DTR-RAGBackbone=LLaMA3.2-3B2026.06 | 36.31 | 29.3 | |
| TrustMarginBackbone=LLaMA3.2-1B2026.06 | 30.6 | 23.2 | |
| DTR-RAGBackbone=LLaMA3.2-1B2026.06 | 30.53 | 23.7 | |
| CLeHe-RAGBackbone=LLaMA3.2-1B2026.06 | 30.49 | 23.2 | |
| BM25-RAGBackbone=LLaMA3.2-3B2026.06 | 30.14 | 23 | |
| IRCoTBackbone=LLaMA3.2-3B2026.06 | 30.03 | 23.9 | |
| FLAREBackbone=LLaMA3.2-1B2026.06 | 28.74 | 21.5 | |
| IRCoTBackbone=LLaMA3.2-1B2026.06 | 27.91 | 21 | |
| DirectBackbone=LLaMA3.2-1B2026.06 | 27.46 | 21.1 | |
| BM25-RAGBackbone=LLaMA3.2-1B2026.06 | 26.01 | 19.7 |