@inproceedings{zhu-etal-2026-task,
title = "Task-Grounded and {EL}2{N}-Informed {RAG} Switch for Domain-Specific Information Extraction",
author = "Zhu, Xingyu and
Bossy, Robert and
Nagy, Balazs and
Vidacs, Laszlo and
N{\'e}dellec, Claire",
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.729/",
doi = "10.18653/2026.findings-emnlp.729",
pages = "15319--15336",
ISBN = "979-8-89176-462-0",
abstract = "Retrieval-augmented generation (RAG) is increasingly popular in information extraction (IE), especially in biological and chemical domains where external knowledge is abundant. Existing RAG-enhanced IE systems, however, typically retrieve for every input, even though retrieval may be unnecessary, introduce distracting evidence, and increase inference cost. To address this issue, we propose RAG Switch, a retrieval decision framework for IE that combines difficulty-aware signals with task-grounded supervision. Specifically, we approximate model-specific training difficulty, quantified by Error L2-Norm (EL2N), from inference-time uncertainty signals, and use it to guide retrieval decisions. We then define retrieval necessity directly from instance-level extraction outcomes without retrieval, yielding task-grounded supervision for learning retrieval decision. Experiments on multiple biological and chemical IE benchmarks across diverse IE backbones show that RAG Switch consistently outperforms representative retrieval decision baselines and achieves a strong efficiency{--}performance trade-off in end-to-end RAG-enhanced IE pipelines."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhu-etal-2026-task">
<titleInfo>
<title>Task-Grounded and EL2N-Informed RAG Switch for Domain-Specific Information Extraction</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xingyu</namePart>
<namePart type="family">Zhu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Robert</namePart>
<namePart type="family">Bossy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Balazs</namePart>
<namePart type="family">Nagy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Laszlo</namePart>
<namePart type="family">Vidacs</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Claire</namePart>
<namePart type="family">Nédellec</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 popular in information extraction (IE), especially in biological and chemical domains where external knowledge is abundant. Existing RAG-enhanced IE systems, however, typically retrieve for every input, even though retrieval may be unnecessary, introduce distracting evidence, and increase inference cost. To address this issue, we propose RAG Switch, a retrieval decision framework for IE that combines difficulty-aware signals with task-grounded supervision. Specifically, we approximate model-specific training difficulty, quantified by Error L2-Norm (EL2N), from inference-time uncertainty signals, and use it to guide retrieval decisions. We then define retrieval necessity directly from instance-level extraction outcomes without retrieval, yielding task-grounded supervision for learning retrieval decision. Experiments on multiple biological and chemical IE benchmarks across diverse IE backbones show that RAG Switch consistently outperforms representative retrieval decision baselines and achieves a strong efficiency–performance trade-off in end-to-end RAG-enhanced IE pipelines.</abstract>
<identifier type="citekey">zhu-etal-2026-task</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.729</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.729/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>15319</start>
<end>15336</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Task-Grounded and EL2N-Informed RAG Switch for Domain-Specific Information Extraction
%A Zhu, Xingyu
%A Bossy, Robert
%A Nagy, Balazs
%A Vidacs, Laszlo
%A Nédellec, Claire
%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 zhu-etal-2026-task
%X Retrieval-augmented generation (RAG) is increasingly popular in information extraction (IE), especially in biological and chemical domains where external knowledge is abundant. Existing RAG-enhanced IE systems, however, typically retrieve for every input, even though retrieval may be unnecessary, introduce distracting evidence, and increase inference cost. To address this issue, we propose RAG Switch, a retrieval decision framework for IE that combines difficulty-aware signals with task-grounded supervision. Specifically, we approximate model-specific training difficulty, quantified by Error L2-Norm (EL2N), from inference-time uncertainty signals, and use it to guide retrieval decisions. We then define retrieval necessity directly from instance-level extraction outcomes without retrieval, yielding task-grounded supervision for learning retrieval decision. Experiments on multiple biological and chemical IE benchmarks across diverse IE backbones show that RAG Switch consistently outperforms representative retrieval decision baselines and achieves a strong efficiency–performance trade-off in end-to-end RAG-enhanced IE pipelines.
%R 10.18653/2026.findings-emnlp.729
%U https://aclanthology.org/2026.findings-emnlp.729/
%U https://doi.org/10.18653/2026.findings-emnlp.729
%P 15319-15336
Markdown (Informal)
[Task-Grounded and EL2N-Informed RAG Switch for Domain-Specific Information Extraction](https://aclanthology.org/2026.findings-emnlp.729/) (Zhu et al., Findings 2026)
ACL