@inproceedings{hao-etal-2026-know,
title = "Know When to Reflect: Entropy-Guided Adaptive Self-Correction for {LLM} Reasoning",
author = "Hao, Xinglei and
Zhao, Haiyu and
Zhang, Chunhong and
Liu, Liping and
Liu, Jiarun and
Hu, Zheng",
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.805/",
doi = "10.18653/2026.findings-emnlp.805",
pages = "17056--17077",
ISBN = "979-8-89176-462-0",
abstract = "Self-correction has emerged as a promising paradigm for enhancing LLM reasoning, yet recent studies reveal a troubling phenomenon: without external oracle feedback, iterative reflection often degrades rather than improves performance. Our analysis uncovers that toxic reflection, where correct answers deteriorate into incorrect ones, occurs at rates comparable to effective corrections. We identify the root cause as the shortsightedness of existing methods: models focus only on the previous output while neglecting the global reasoning state, rendering them unable to quantify uncertainty. To address this, we propose two complementary reflection signals {---} Reflecting Entropy for snapshot disagreement among reasoning trajectories and Belief Shift for inter-round belief revision {---} paired with a two-stage hierarchical controller that schedules reflection adaptively. We instantiate these ideas as REGAR (Reflecting Entropy-Guided Adaptive Reflection), a framework that maintains a pool of candidate reasoning trajectories and applies a single, uniform reflection primitive when correction is needed. Experiments on mathematical and commonsense reasoning demonstrate consistent improvements over strong baselines while reducing computational overhead, highlighting both effectiveness and efficiency."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="hao-etal-2026-know">
<titleInfo>
<title>Know When to Reflect: Entropy-Guided Adaptive Self-Correction for LLM Reasoning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xinglei</namePart>
<namePart type="family">Hao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haiyu</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chunhong</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Liping</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiarun</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zheng</namePart>
<namePart type="family">Hu</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>Self-correction has emerged as a promising paradigm for enhancing LLM reasoning, yet recent studies reveal a troubling phenomenon: without external oracle feedback, iterative reflection often degrades rather than improves performance. Our analysis uncovers that toxic reflection, where correct answers deteriorate into incorrect ones, occurs at rates comparable to effective corrections. We identify the root cause as the shortsightedness of existing methods: models focus only on the previous output while neglecting the global reasoning state, rendering them unable to quantify uncertainty. To address this, we propose two complementary reflection signals — Reflecting Entropy for snapshot disagreement among reasoning trajectories and Belief Shift for inter-round belief revision — paired with a two-stage hierarchical controller that schedules reflection adaptively. We instantiate these ideas as REGAR (Reflecting Entropy-Guided Adaptive Reflection), a framework that maintains a pool of candidate reasoning trajectories and applies a single, uniform reflection primitive when correction is needed. Experiments on mathematical and commonsense reasoning demonstrate consistent improvements over strong baselines while reducing computational overhead, highlighting both effectiveness and efficiency.</abstract>
<identifier type="citekey">hao-etal-2026-know</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.805</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.805/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>17056</start>
<end>17077</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Know When to Reflect: Entropy-Guided Adaptive Self-Correction for LLM Reasoning
%A Hao, Xinglei
%A Zhao, Haiyu
%A Zhang, Chunhong
%A Liu, Liping
%A Liu, Jiarun
%A Hu, Zheng
%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 hao-etal-2026-know
%X Self-correction has emerged as a promising paradigm for enhancing LLM reasoning, yet recent studies reveal a troubling phenomenon: without external oracle feedback, iterative reflection often degrades rather than improves performance. Our analysis uncovers that toxic reflection, where correct answers deteriorate into incorrect ones, occurs at rates comparable to effective corrections. We identify the root cause as the shortsightedness of existing methods: models focus only on the previous output while neglecting the global reasoning state, rendering them unable to quantify uncertainty. To address this, we propose two complementary reflection signals — Reflecting Entropy for snapshot disagreement among reasoning trajectories and Belief Shift for inter-round belief revision — paired with a two-stage hierarchical controller that schedules reflection adaptively. We instantiate these ideas as REGAR (Reflecting Entropy-Guided Adaptive Reflection), a framework that maintains a pool of candidate reasoning trajectories and applies a single, uniform reflection primitive when correction is needed. Experiments on mathematical and commonsense reasoning demonstrate consistent improvements over strong baselines while reducing computational overhead, highlighting both effectiveness and efficiency.
%R 10.18653/2026.findings-emnlp.805
%U https://aclanthology.org/2026.findings-emnlp.805/
%U https://doi.org/10.18653/2026.findings-emnlp.805
%P 17056-17077
Markdown (Informal)
[Know When to Reflect: Entropy-Guided Adaptive Self-Correction for LLM Reasoning](https://aclanthology.org/2026.findings-emnlp.805/) (Hao et al., Findings 2026)
ACL