RAG Utility Evaluation on FEVER
77Hit RatePromptGuard
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PromptGuardDefense=PromptGuard2026.05 | 77 | 66.1 | 75.5 | |
| No DefenseDefense=No Defense2026.05 | 72.8 | 62.5 | 75 | |
| MetaSecAlignDefense=MetaSecAlign2026.05 | 72.8 | 62.5 | 74.9 | |
| CleanBaseDefense=CleanBase2026.05 | 71.4 | 61 | 75.1 | |
| CleanBase + MetaSecAlignDefense=CleanBase + MetaSecAlign2026.05 | 71.4 | 61 | 77.9 | |
| TrustRAGDefense=TrustRAG2026.05 | 57.6 | 49.3 | 70.9 |