@inproceedings{won-etal-2026-sieve,
title = "Sieve and Sage: Efficient Distraction Filtering for Reliable {RALM} Abstention",
author = "Won, Jongbin and
AN, SungGeun and
Lee, Jay-Yoon",
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.2356/",
doi = "10.18653/2026.findings-emnlp.2356",
pages = "50655--50674",
ISBN = "979-8-89176-462-0",
abstract = "Just as Socrates recognized the limits of his own knowledge, Retrieval-Augmented Language Models (RALMs) should learn to \textit{abstain} when the retrieved evidence cannot support a reliable response. Existing approaches largely rely on monolithic LLMs to handle heterogeneous retrieval failures in a single step, resulting in limited abstention performance and high computational costs. We instead decompose retrieval failures into two distinct states: (i) the \textit{unanswerable state}, where the required evidence is absent, and (ii) the \textit{distracted state}, where relevant evidence is mixed with conflicting, negated, or adversarial information. Based on this decomposition, we introduce a lightweight module (\textbf{Sieve}) that screens retrieved document sets for distracting evidence before invoking a costly LLM (\textbf{Sage}) for grounded generation and abstention. Evaluated across both general and high-stakes expert domains, our Sieve and Sage framework preemptively detects distracting noise, improving system accuracy by up to 69.4{\%}p and Macro-F1 by 55.2{\%}p compared to one-stage baselines. Furthermore, it achieves up to a 1.99$\times$ speedup, establishing a highly efficient and reliable abstention pipeline for RALM with abstention."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="won-etal-2026-sieve">
<titleInfo>
<title>Sieve and Sage: Efficient Distraction Filtering for Reliable RALM Abstention</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jongbin</namePart>
<namePart type="family">Won</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">SungGeun</namePart>
<namePart type="family">AN</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jay-Yoon</namePart>
<namePart type="family">Lee</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>Just as Socrates recognized the limits of his own knowledge, Retrieval-Augmented Language Models (RALMs) should learn to abstain when the retrieved evidence cannot support a reliable response. Existing approaches largely rely on monolithic LLMs to handle heterogeneous retrieval failures in a single step, resulting in limited abstention performance and high computational costs. We instead decompose retrieval failures into two distinct states: (i) the unanswerable state, where the required evidence is absent, and (ii) the distracted state, where relevant evidence is mixed with conflicting, negated, or adversarial information. Based on this decomposition, we introduce a lightweight module (Sieve) that screens retrieved document sets for distracting evidence before invoking a costly LLM (Sage) for grounded generation and abstention. Evaluated across both general and high-stakes expert domains, our Sieve and Sage framework preemptively detects distracting noise, improving system accuracy by up to 69.4%p and Macro-F1 by 55.2%p compared to one-stage baselines. Furthermore, it achieves up to a 1.99\times speedup, establishing a highly efficient and reliable abstention pipeline for RALM with abstention.</abstract>
<identifier type="citekey">won-etal-2026-sieve</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.2356</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2356/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>50655</start>
<end>50674</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Sieve and Sage: Efficient Distraction Filtering for Reliable RALM Abstention
%A Won, Jongbin
%A AN, SungGeun
%A Lee, Jay-Yoon
%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 won-etal-2026-sieve
%X Just as Socrates recognized the limits of his own knowledge, Retrieval-Augmented Language Models (RALMs) should learn to abstain when the retrieved evidence cannot support a reliable response. Existing approaches largely rely on monolithic LLMs to handle heterogeneous retrieval failures in a single step, resulting in limited abstention performance and high computational costs. We instead decompose retrieval failures into two distinct states: (i) the unanswerable state, where the required evidence is absent, and (ii) the distracted state, where relevant evidence is mixed with conflicting, negated, or adversarial information. Based on this decomposition, we introduce a lightweight module (Sieve) that screens retrieved document sets for distracting evidence before invoking a costly LLM (Sage) for grounded generation and abstention. Evaluated across both general and high-stakes expert domains, our Sieve and Sage framework preemptively detects distracting noise, improving system accuracy by up to 69.4%p and Macro-F1 by 55.2%p compared to one-stage baselines. Furthermore, it achieves up to a 1.99\times speedup, establishing a highly efficient and reliable abstention pipeline for RALM with abstention.
%R 10.18653/2026.findings-emnlp.2356
%U https://aclanthology.org/2026.findings-emnlp.2356/
%U https://doi.org/10.18653/2026.findings-emnlp.2356
%P 50655-50674
Markdown (Informal)
[Sieve and Sage: Efficient Distraction Filtering for Reliable RALM Abstention](https://aclanthology.org/2026.findings-emnlp.2356/) (Won et al., Findings 2026)
ACL