False Premise Detection on FEVER
0.961TP RateWRAG (with RAG)
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| WRAG (with RAG)Retrieval input format=Orig2025.04 | 0.961 | 0.612 | 0.388 | 0.039 | 0.824 | 0.79 | |
| WRAG (with RAG)Retrieval input format=LF-Retr2025.04 | 0.941 | 0.837 | 0.163 | 0.059 | 0.897 | 0.89 | |
| WRAG (with RAG)Retrieval input format=LF-Both2025.04 | 0.941 | 0.837 | 0.163 | 0.059 | 0.897 | 0.89 | |
| GraphRAG/ToG2025.04 | 0.882 | 0.837 | 0.163 | 0.118 | 0.865 | 0.86 | |
| Direct2025.04 | 0.188 | 0.985 | 0.015 | 0.812 | 0.313 | 0.584 |