List your company

End-to-End Defense in RAG on SciFact

0ASRTrustRAG
-3.9622.7749.576.23May 1, 2026May 5, 2026May 9, 2026May 13, 2026May 17, 2026May 21, 2026May 26, 2026

Evaluation results

MethodLinks
TrustRAGAttack Scenario=GL2026.0500
PromptGuardAttack Scenario=GL2026.0501
CleanBaseAttack Scenario=GL2026.0500
CleanBase + MetaSecAlignAttack Scenario=GL2026.0500
Cordon-MASSeed=422026.052—
CleanBaseAttack Scenario=PT2026.0530.4
CleanBase + MetaSecAlignAttack Scenario=PT2026.0530.4
PromptGuardAttack Scenario=FC2026.0540
CleanBaseAttack Scenario=MA2026.0540
PromptGuardAttack Scenario=NA2026.0550
CleanBaseAttack Scenario=CI2026.0550
CleanBaseAttack Scenario=FC2026.0560
PromptGuardAttack Scenario=MA2026.0570
CleanBaseAttack Scenario=PB2026.0570
CleanBase + MetaSecAlignAttack Scenario=NA2026.0570
CleanBase + MetaSecAlignAttack Scenario=FC2026.0570
CleanBase + MetaSecAlignAttack Scenario=MA2026.0570
CleanBase + MetaSecAlignAttack Scenario=PB2026.0570
CleanBaseAttack Scenario=NA2026.0580
CleanBase + MetaSecAlignAttack Scenario=CI2026.0580
CleanBase + MetaSecAlignAttack Scenario=CA2026.0580
PromptGuardAttack Scenario=CI2026.0590
CleanBaseAttack Scenario=CA2026.0590
PromptGuardAttack Scenario=CA2026.05100
TrustRAGAttack Scenario=MA2026.05120
CleanBase + MetaSecAlignAttack Scenario=PW2026.05125.4
TrustRAGAttack Scenario=NA2026.0514100
CleanBaseAttack Scenario=PW2026.05155.4
TrustRAGAttack Scenario=CI2026.05160
TrustRAGAttack Scenario=PB2026.05160
TrustRAGAttack Scenario=FC2026.05170
TrustRAGAttack Scenario=PT2026.05170
TrustRAGAttack Scenario=PW2026.05180.4
TrustRAGAttack Scenario=CA2026.05200
MetaSecAlignAttack Scenario=FC2026.0521100
DataSentinelAttack Scenario=CA2026.05219.6
PromptGuardAttack Scenario=PT2026.052233.2
MetaSecAlignAttack Scenario=NA2026.0523100
MetaSecAlignAttack Scenario=PB2026.0523100
MetaSecAlignAttack Scenario=MA2026.0526100
MetaSecAlignAttack Scenario=CI2026.0529100
MetaSecAlignAttack Scenario=PW2026.0529100
MetaSecAlignAttack Scenario=CA2026.0530100
MetaSecAlignAttack Scenario=PT2026.053098.8
DataSentinelAttack Scenario=PT2026.053398.8
No DefenseAttack Scenario=PT2026.053498.8
DataSentinelAttack Scenario=CI2026.053820.4
DataSentinelAttack Scenario=FC2026.053818.8
DebateSeed=422026.0538—
DataSentinelAttack Scenario=MA2026.054018.6
RobustRAGSeed=422026.0544—
PromptGuardAttack Scenario=PB2026.054762.4
No DefenseAttack Scenario=PB2026.0548100
PromptGuardAttack Scenario=PW2026.055470.2
MetaSecAlignAttack Scenario=GL2026.0555100
DataSentinelAttack Scenario=PB2026.0556100
ParaphraseSeed=422026.0558—
TrustRAGSeed=422026.0560—
No DefenseAttack Scenario=PW2026.0562100
Vanilla RAGSeed=422026.0562—
DataSentinelAttack Scenario=NA2026.056566
DataSentinelAttack Scenario=PW2026.056799.2
DataSentinelAttack Scenario=GL2026.058051
No DefenseAttack Scenario=FC2026.0589100
No DefenseAttack Scenario=GL2026.0590100
No DefenseAttack Scenario=CA2026.0597100
No DefenseAttack Scenario=NA2026.0599100
No DefenseAttack Scenario=CI2026.0599100
No DefenseAttack Scenario=MA2026.0599100