End-to-End Defense in RAG on FEVER
0ASRTrustRAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TrustRAGAttack Scenario=GL2026.05 | 0 | 0 | |
| CleanBaseAttack Scenario=GL2026.05 | 0 | 0 | |
| CleanBase + MetaSecAlignAttack Scenario=GL2026.05 | 0 | 0 | |
| PromptGuardAttack Scenario=FC2026.05 | 3 | 0 | |
| PromptGuardAttack Scenario=CA2026.05 | 3 | 0 | |
| PromptGuardAttack Scenario=NA2026.05 | 4 | 0 | |
| PromptGuardAttack Scenario=MA2026.05 | 4 | 0 | |
| PromptGuardAttack Scenario=GL2026.05 | 5 | 2 | |
| CleanBaseAttack Scenario=NA2026.05 | 5 | 0 | |
| CleanBaseAttack Scenario=CI2026.05 | 5 | 0.2 | |
| CleanBaseAttack Scenario=CA2026.05 | 5 | 0 | |
| CleanBase + MetaSecAlignAttack Scenario=NA2026.05 | 5 | 0 | |
| PromptGuardAttack Scenario=CI2026.05 | 6 | 0 | |
| CleanBaseAttack Scenario=PB2026.05 | 6 | 12.4 | |
| CleanBase + MetaSecAlignAttack Scenario=PB2026.05 | 6 | 12.4 | |
| CleanBase + MetaSecAlignAttack Scenario=CI2026.05 | 7 | 0.2 | |
| CleanBase + MetaSecAlignAttack Scenario=FC2026.05 | 8 | 3.6 | |
| CleanBase + MetaSecAlignAttack Scenario=CA2026.05 | 8 | 0 | |
| CleanBase + MetaSecAlignAttack Scenario=MA2026.05 | 10 | 5.2 | |
| CleanBaseAttack Scenario=PT2026.05 | 11 | 30.6 | |
| CleanBase + MetaSecAlignAttack Scenario=PT2026.05 | 11 | 30.6 | |
| CleanBaseAttack Scenario=FC2026.05 | 12 | 3.6 | |
| PromptGuardAttack Scenario=PT2026.05 | 13 | 31.8 | |
| CleanBaseAttack Scenario=MA2026.05 | 13 | 5.2 | |
| TrustRAGAttack Scenario=PT2026.05 | 14 | 1.8 | |
| DataSentinelAttack Scenario=PT2026.05 | 15 | 95.8 | |
| MetaSecAlignAttack Scenario=FC2026.05 | 16 | 81.2 | |
| TrustRAGAttack Scenario=CI2026.05 | 16 | 1 | |
| No DefenseAttack Scenario=PT2026.05 | 17 | 96 | |
| TrustRAGAttack Scenario=MA2026.05 | 17 | 0.6 | |
| TrustRAGAttack Scenario=PB2026.05 | 17 | 0.4 | |
| TrustRAGAttack Scenario=NA2026.05 | 18 | 93.2 | |
| TrustRAGAttack Scenario=FC2026.05 | 18 | 2 | |
| PromptGuardAttack Scenario=PB2026.05 | 18 | 30.8 | |
| MetaSecAlignAttack Scenario=PT2026.05 | 19 | 96 | |
| TrustRAGAttack Scenario=CA2026.05 | 19 | 2 | |
| TrustRAGAttack Scenario=PW2026.05 | 20 | 3.2 | |
| CleanBase + MetaSecAlignAttack Scenario=PW2026.05 | 20 | 65.2 | |
| MetaSecAlignAttack Scenario=PB2026.05 | 21 | 99.2 | |
| DataSentinelAttack Scenario=CA2026.05 | 21 | 12.4 | |
| MetaSecAlignAttack Scenario=PW2026.05 | 22 | 97.2 | |
| MetaSecAlignAttack Scenario=NA2026.05 | 23 | 97.8 | |
| PromptGuardAttack Scenario=PW2026.05 | 24 | 50 | |
| MetaSecAlignAttack Scenario=CA2026.05 | 26 | 86 | |
| DataSentinelAttack Scenario=PB2026.05 | 26 | 99.2 | |
| DataSentinelAttack Scenario=FC2026.05 | 27 | 14.4 | |
| No DefenseAttack Scenario=PB2026.05 | 28 | 99.2 | |
| MetaSecAlignAttack Scenario=MA2026.05 | 28 | 88.8 | |
| CleanBaseAttack Scenario=PW2026.05 | 30 | 65.2 | |
| DataSentinelAttack Scenario=CI2026.05 | 33 | 18.4 | |
| DataSentinelAttack Scenario=MA2026.05 | 33 | 17.2 | |
| DataSentinelAttack Scenario=PW2026.05 | 36 | 96.8 | |
| No DefenseAttack Scenario=PW2026.05 | 37 | 97.2 | |
| MetaSecAlignAttack Scenario=CI2026.05 | 40 | 91.8 | |
| MetaSecAlignAttack Scenario=GL2026.05 | 45 | 100 | |
| DataSentinelAttack Scenario=GL2026.05 | 55 | 47 | |
| DataSentinelAttack Scenario=NA2026.05 | 59 | 60 | |
| No DefenseAttack Scenario=FC2026.05 | 73 | 81.2 | |
| No DefenseAttack Scenario=CA2026.05 | 89 | 86 | |
| No DefenseAttack Scenario=MA2026.05 | 92 | 88.8 | |
| No DefenseAttack Scenario=CI2026.05 | 95 | 91.8 | |
| No DefenseAttack Scenario=GL2026.05 | 95 | 100 | |
| No DefenseAttack Scenario=NA2026.05 | 97 | 97.8 |