Introduces NPAS and AV Filter using LLM attention weights to defend RAG against poisoning, reporting up to 20% accuracy gains while adaptive attacks reach 35% success.
Attacking open-domain question answering by injecting misinformation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Through the Stealth Lens: Attention-Aware Defenses Against Poisoning in RAG
Introduces NPAS and AV Filter using LLM attention weights to defend RAG against poisoning, reporting up to 20% accuracy gains while adaptive attacks reach 35% success.