RAGCharacter localizes poisoned character spans in RAG evidence via prompt-conditioned counterfactual masking and achieves the best accuracy-over-attribution trade-off across tested attacks and models.
In: Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing
2 Pith papers cite this work, alongside 12 external citations. Polarity classification is still indexing.
2
Pith papers citing it
12
external citations · OpenAlex
citation-role summary
background 2
citation-polarity summary
fields
cs.CR 2years
2026 2verdicts
UNVERDICTED 2roles
background 2polarities
background 2representative citing papers
Anonymization placement in RAG—at the dataset or at the generated answer—creates observable differences in privacy protection versus response utility.
citing papers explorer
-
Needle-in-RAG: Prompt-Conditioned Character-Level Traceback of Poisoned Spans in Retrieved Evidence
RAGCharacter localizes poisoned character spans in RAG evidence via prompt-conditioned counterfactual masking and achieves the best accuracy-over-attribution trade-off across tested attacks and models.
-
A Case Study on the Impact of Anonymization Along the RAG Pipeline
Anonymization placement in RAG—at the dataset or at the generated answer—creates observable differences in privacy protection versus response utility.