<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lee-etal-2026-reflection">
    <titleInfo>
        <title>When Reflection Helps: Utility-Guided Advantage Rescaling for Efficient Reasoning</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Jungyu</namePart>
        <namePart type="family">Lee</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Kunhui</namePart>
        <namePart type="family">Lee</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Dongwook</namePart>
        <namePart type="family">Min</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jeonhwi</namePart>
        <namePart type="family">Kang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Seung-Hoon</namePart>
        <namePart type="family">Na</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>Large reasoning models have recently demonstrated emergent “aha moments” that enable reflective reasoning. However, a non-trivial portion of reflective steps are “low-utility,” where reasoning errors are not corrected, or even correct reasoning trajectories are shifted toward incorrect ones. To address low-utility reflective continuation, this paper proposes RUAR, a utility-guided advantage rescaling method for reflective reasoning that estimates step-level utility from changes in correct-answer probability, without human process labels or a separately learned process reward model. RUAR augments outcome-supervised RL by rescaling the token-level advantages within each reflective reasoning step according to the estimated utility of the reflective continuation, rather than applying a uniform trajectory-level advantage across the reasoning trace. The utility is estimated through answer-forcing utility estimation, which measures how the model’s correct-answer probability changes before and after the reflective step by forcing answer generation from both reasoning states using answer-forcing prompts. Furthermore, RUAR introduces an overthinking-mitigating rescaling mechanism, where utility-guided rescaling is no longer applied from the earliest answer-ready reflection step onward. Experiments demonstrate that RUAR improves the accuracy–efficiency trade-off, substantially reducing generated token usage while often preserving or improving final-answer accuracy.</abstract>
    <identifier type="citekey">lee-etal-2026-reflection</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.1441/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>30739</start>
            <end>30760</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T When Reflection Helps: Utility-Guided Advantage Rescaling for Efficient Reasoning
%A Lee, Jungyu
%A Lee, Kunhui
%A Min, Dongwook
%A Kang, Jeonhwi
%A Na, Seung-Hoon
%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 lee-etal-2026-reflection
%X Large reasoning models have recently demonstrated emergent “aha moments” that enable reflective reasoning. However, a non-trivial portion of reflective steps are “low-utility,” where reasoning errors are not corrected, or even correct reasoning trajectories are shifted toward incorrect ones. To address low-utility reflective continuation, this paper proposes RUAR, a utility-guided advantage rescaling method for reflective reasoning that estimates step-level utility from changes in correct-answer probability, without human process labels or a separately learned process reward model. RUAR augments outcome-supervised RL by rescaling the token-level advantages within each reflective reasoning step according to the estimated utility of the reflective continuation, rather than applying a uniform trajectory-level advantage across the reasoning trace. The utility is estimated through answer-forcing utility estimation, which measures how the model’s correct-answer probability changes before and after the reflective step by forcing answer generation from both reasoning states using answer-forcing prompts. Furthermore, RUAR introduces an overthinking-mitigating rescaling mechanism, where utility-guided rescaling is no longer applied from the earliest answer-ready reflection step onward. Experiments demonstrate that RUAR improves the accuracy–efficiency trade-off, substantially reducing generated token usage while often preserving or improving final-answer accuracy.
%U https://aclanthology.org/2026.findings-emnlp.1441/
%P 30739-30760

Markdown (Informal)

[When Reflection Helps: Utility-Guided Advantage Rescaling for Efficient Reasoning](https://aclanthology.org/2026.findings-emnlp.1441/) (Lee et al., Findings 2026)

• When Reflection Helps: Utility-Guided Advantage Rescaling for Efficient Reasoning (Lee et al., Findings 2026)

ACL

• Jungyu Lee, Kunhui Lee, Dongwook Min, Jeonhwi Kang, and Seung-Hoon Na. 2026. When Reflection Helps: Utility-Guided Advantage Rescaling for Efficient Reasoning. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 30739–30760, Budapest, Hungary. Association for Computational Linguistics.