@inproceedings{huang-etal-2026-confrag,
title = "{CONFRAG}: Confidence-Guided Retrieval-Augmented Generation",
author = "Huang, Yin and
Xu, Yifan Ethan and
Sun, Kai and
Yan, Vera and
Sun, Alicia Yi and
Khan, Haidar and
Nguyen, Jimmy and
Chen, Jingxiang and
Kachuee, Mohammad and
Lin, Zhaojiang and
Liu, Yue and
Colak, Aaron and
Kumar, Anuj and
Yih, Wen-tau and
Dong, Xin Luna",
editor = "Atanasova, Pepa and
Haffari, Gholamreza and
Wang, Minghan",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-industry.171/",
pages = "2905--2923",
ISBN = "979-8-89176-460-6",
abstract = "Production Retrieval-Augmented Generation (RAG) pays retrieval latency and serving cost on every query {--} even when the model already knows the answer. We present CONFRAG, a confidence-guided strategy that invokes RAG only when a large language model (LLM) is unlikely to be correct, cutting external retrievals by 5-39{\%} and tail latency by up to 800ms (P90) in a real RAG system while matching the answer quality of always-on RAG. CONFRAG is enabled by CONFQA, a lightweight fine-tuning strategy (one epoch over 3K knowledge-graph facts) that teaches an LLM to answer only when confident and otherwise say ``I am unsure'', cutting hallucination rates from 20-40{\%} to 5{\%} or below across multiple factuality benchmarks. Two choices make it work in practice: a dampening prompt that discourages overconfident hallucinations, and training on atomic facts that generalizes confidence calibration to unseen domains and question types. At serving time, LLM-QA and RAG run in parallel with early-stopping, so confident answers return immediately."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huang-etal-2026-confrag">
<titleInfo>
<title>CONFRAG: Confidence-Guided Retrieval-Augmented Generation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yin</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yifan</namePart>
<namePart type="given">Ethan</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kai</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vera</namePart>
<namePart type="family">Yan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alicia</namePart>
<namePart type="given">Yi</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haidar</namePart>
<namePart type="family">Khan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jimmy</namePart>
<namePart type="family">Nguyen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jingxiang</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="family">Kachuee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhaojiang</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aaron</namePart>
<namePart type="family">Colak</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anuj</namePart>
<namePart type="family">Kumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wen-tau</namePart>
<namePart type="family">Yih</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xin</namePart>
<namePart type="given">Luna</namePart>
<namePart type="family">Dong</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pepa</namePart>
<namePart type="family">Atanasova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gholamreza</namePart>
<namePart type="family">Haffari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Minghan</namePart>
<namePart type="family">Wang</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">979-8-89176-460-6</identifier>
</relatedItem>
<abstract>Production Retrieval-Augmented Generation (RAG) pays retrieval latency and serving cost on every query – even when the model already knows the answer. We present CONFRAG, a confidence-guided strategy that invokes RAG only when a large language model (LLM) is unlikely to be correct, cutting external retrievals by 5-39% and tail latency by up to 800ms (P90) in a real RAG system while matching the answer quality of always-on RAG. CONFRAG is enabled by CONFQA, a lightweight fine-tuning strategy (one epoch over 3K knowledge-graph facts) that teaches an LLM to answer only when confident and otherwise say “I am unsure”, cutting hallucination rates from 20-40% to 5% or below across multiple factuality benchmarks. Two choices make it work in practice: a dampening prompt that discourages overconfident hallucinations, and training on atomic facts that generalizes confidence calibration to unseen domains and question types. At serving time, LLM-QA and RAG run in parallel with early-stopping, so confident answers return immediately.</abstract>
<identifier type="citekey">huang-etal-2026-confrag</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-industry.171/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>2905</start>
<end>2923</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CONFRAG: Confidence-Guided Retrieval-Augmented Generation
%A Huang, Yin
%A Xu, Yifan Ethan
%A Sun, Kai
%A Yan, Vera
%A Sun, Alicia Yi
%A Khan, Haidar
%A Nguyen, Jimmy
%A Chen, Jingxiang
%A Kachuee, Mohammad
%A Lin, Zhaojiang
%A Liu, Yue
%A Colak, Aaron
%A Kumar, Anuj
%A Yih, Wen-tau
%A Dong, Xin Luna
%Y Atanasova, Pepa
%Y Haffari, Gholamreza
%Y Wang, Minghan
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-460-6
%F huang-etal-2026-confrag
%X Production Retrieval-Augmented Generation (RAG) pays retrieval latency and serving cost on every query – even when the model already knows the answer. We present CONFRAG, a confidence-guided strategy that invokes RAG only when a large language model (LLM) is unlikely to be correct, cutting external retrievals by 5-39% and tail latency by up to 800ms (P90) in a real RAG system while matching the answer quality of always-on RAG. CONFRAG is enabled by CONFQA, a lightweight fine-tuning strategy (one epoch over 3K knowledge-graph facts) that teaches an LLM to answer only when confident and otherwise say “I am unsure”, cutting hallucination rates from 20-40% to 5% or below across multiple factuality benchmarks. Two choices make it work in practice: a dampening prompt that discourages overconfident hallucinations, and training on atomic facts that generalizes confidence calibration to unseen domains and question types. At serving time, LLM-QA and RAG run in parallel with early-stopping, so confident answers return immediately.
%U https://aclanthology.org/2026.emnlp-industry.171/
%P 2905-2923
Markdown (Informal)
[CONFRAG: Confidence-Guided Retrieval-Augmented Generation](https://aclanthology.org/2026.emnlp-industry.171/) (Huang et al., EMNLP 2026)
ACL
- Yin Huang, Yifan Ethan Xu, Kai Sun, Vera Yan, Alicia Yi Sun, Haidar Khan, Jimmy Nguyen, Jingxiang Chen, Mohammad Kachuee, Zhaojiang Lin, Yue Liu, Aaron Colak, Anuj Kumar, Wen-tau Yih, and Xin Luna Dong. 2026. CONFRAG: Confidence-Guided Retrieval-Augmented Generation. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 2905–2923, Budapest, Hungary. Association for Computational Linguistics.