@inproceedings{che-etal-2026-towards,
title = "Towards Robust Retrieval-Augmented Generation via Reliability-Aware Adaptive Evidence Selection",
author = "Che, Minkyu and
Oh, Jihwan and
Choi, Jeonghwan and
Jung, Jeesu and
Song, Hwanjun",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.2302/",
doi = "10.18653/2026.findings-emnlp.2302",
pages = "49353--49374",
ISBN = "979-8-89176-462-0",
abstract = "Retrieval-augmented generation (RAG) is increasingly deployed on unverified corpora, where an adversary can inject poisoned documents to steer the model toward an attacker-specified answer. Existing defenses rely on auxiliary heuristics and break down once their assumptions about the attack type or poisoning level are violated. We propose SAFER-RAG, a defense that derives its reliability signal from the adversarial objective itself. Since poisoned documents are explicitly optimized to rank highly for the target query, they remain anchored at top ranks under semantically equivalent rewrites, while clean documents shift naturally. SAFER-RAG captures this rank-stability gap through spectral analysis, partitions retrieved chunks into Safe and Unsafe Sets, and adaptively selects evidence under a safe-first strategy that avoids over-filtering useful evidence on clean or lightly poisoned corpora. Across three QA benchmarks, three attack types, and four poisoning levels, SAFER-RAG consistently achieves the highest accuracy and the lowest attack success rate. The code will be released upon acceptance."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="che-etal-2026-towards">
<titleInfo>
<title>Towards Robust Retrieval-Augmented Generation via Reliability-Aware Adaptive Evidence Selection</title>
</titleInfo>
<name type="personal">
<namePart type="given">Minkyu</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jihwan</namePart>
<namePart type="family">Oh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jeonghwan</namePart>
<namePart type="family">Choi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jeesu</namePart>
<namePart type="family">Jung</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hwanjun</namePart>
<namePart type="family">Song</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Retrieval-augmented generation (RAG) is increasingly deployed on unverified corpora, where an adversary can inject poisoned documents to steer the model toward an attacker-specified answer. Existing defenses rely on auxiliary heuristics and break down once their assumptions about the attack type or poisoning level are violated. We propose SAFER-RAG, a defense that derives its reliability signal from the adversarial objective itself. Since poisoned documents are explicitly optimized to rank highly for the target query, they remain anchored at top ranks under semantically equivalent rewrites, while clean documents shift naturally. SAFER-RAG captures this rank-stability gap through spectral analysis, partitions retrieved chunks into Safe and Unsafe Sets, and adaptively selects evidence under a safe-first strategy that avoids over-filtering useful evidence on clean or lightly poisoned corpora. Across three QA benchmarks, three attack types, and four poisoning levels, SAFER-RAG consistently achieves the highest accuracy and the lowest attack success rate. The code will be released upon acceptance.</abstract>
<identifier type="citekey">che-etal-2026-towards</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.2302</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2302/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>49353</start>
<end>49374</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Towards Robust Retrieval-Augmented Generation via Reliability-Aware Adaptive Evidence Selection
%A Che, Minkyu
%A Oh, Jihwan
%A Choi, Jeonghwan
%A Jung, Jeesu
%A Song, Hwanjun
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F che-etal-2026-towards
%X Retrieval-augmented generation (RAG) is increasingly deployed on unverified corpora, where an adversary can inject poisoned documents to steer the model toward an attacker-specified answer. Existing defenses rely on auxiliary heuristics and break down once their assumptions about the attack type or poisoning level are violated. We propose SAFER-RAG, a defense that derives its reliability signal from the adversarial objective itself. Since poisoned documents are explicitly optimized to rank highly for the target query, they remain anchored at top ranks under semantically equivalent rewrites, while clean documents shift naturally. SAFER-RAG captures this rank-stability gap through spectral analysis, partitions retrieved chunks into Safe and Unsafe Sets, and adaptively selects evidence under a safe-first strategy that avoids over-filtering useful evidence on clean or lightly poisoned corpora. Across three QA benchmarks, three attack types, and four poisoning levels, SAFER-RAG consistently achieves the highest accuracy and the lowest attack success rate. The code will be released upon acceptance.
%R 10.18653/2026.findings-emnlp.2302
%U https://aclanthology.org/2026.findings-emnlp.2302/
%U https://doi.org/10.18653/2026.findings-emnlp.2302
%P 49353-49374
Markdown (Informal)
[Towards Robust Retrieval-Augmented Generation via Reliability-Aware Adaptive Evidence Selection](https://aclanthology.org/2026.findings-emnlp.2302/) (Che et al., Findings 2026)
ACL