<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-2025-confidence">
    <titleInfo>
        <title>Confidence-Aware Reasoning: Optimizing Self-Guided Thinking Trajectories in Large Reasoning Models</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Jiaxin</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2025-11</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Saloni</namePart>
            <namePart type="family">Potdar</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Lina</namePart>
            <namePart type="family">Rojas-Barahona</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Sebastien</namePart>
            <namePart type="family">Montella</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Suzhou (China)</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-333-3</identifier>
    </relatedItem>
    <abstract>Chain-of-thought enables large reasoning models (LRMs) to reason through multi-step problems but often leads to unnecessarily long or redundant reasoning traces, a phenomenon known as overthinking. This results in inflated inference costs and potential degradation in answer quality. To address these challenges, we propose Confidence-Aware Reasoning (), an inference-time framework that optimizes reasoning trajectories by selectively pruning low-utility reasoning blocks and halting early when sufficient confidence has been achieved. is theoretically grounded in Bayesian optimal experimental design, treating each reasoning block as a sequential decision whose utility is approximated by its marginal contribution to reducing final answer uncertainty. We introduce a lightweight implementation that leverages token-level confidence to dynamically modulate reasoning depth without additional supervision. Evaluations on multiple benchmarks, including AMC, AIME, GPQA-Diamond, and MATH-500 show that improves answer accuracy by up to +13.3%, while reducing average reasoning length by 40%–50%. Our findings demonstrate that information-theoretic insights can effectively control self-guided reasoning and enable LRMs to “think just enough” at test time.</abstract>
    <identifier type="citekey">zhang-2025-confidence</identifier>
    <location>
        <url>https://aclanthology.org/2025.emnlp-industry.146/</url>
    </location>
    <part>
        <date>2025-11</date>
        <extent unit="page">
            <start>2081</start>
            <end>2095</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Confidence-Aware Reasoning: Optimizing Self-Guided Thinking Trajectories in Large Reasoning Models
%A Zhang, Jiaxin
%Y Potdar, Saloni
%Y Rojas-Barahona, Lina
%Y Montella, Sebastien
%S Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou (China)
%@ 979-8-89176-333-3
%F zhang-2025-confidence
%X Chain-of-thought enables large reasoning models (LRMs) to reason through multi-step problems but often leads to unnecessarily long or redundant reasoning traces, a phenomenon known as overthinking. This results in inflated inference costs and potential degradation in answer quality. To address these challenges, we propose Confidence-Aware Reasoning (), an inference-time framework that optimizes reasoning trajectories by selectively pruning low-utility reasoning blocks and halting early when sufficient confidence has been achieved. is theoretically grounded in Bayesian optimal experimental design, treating each reasoning block as a sequential decision whose utility is approximated by its marginal contribution to reducing final answer uncertainty. We introduce a lightweight implementation that leverages token-level confidence to dynamically modulate reasoning depth without additional supervision. Evaluations on multiple benchmarks, including AMC, AIME, GPQA-Diamond, and MATH-500 show that improves answer accuracy by up to +13.3%, while reducing average reasoning length by 40%–50%. Our findings demonstrate that information-theoretic insights can effectively control self-guided reasoning and enable LRMs to “think just enough” at test time.
%U https://aclanthology.org/2025.emnlp-industry.146/
%P 2081-2095

Markdown (Informal)

[Confidence-Aware Reasoning: Optimizing Self-Guided Thinking Trajectories in Large Reasoning Models](https://aclanthology.org/2025.emnlp-industry.146/) (Zhang, EMNLP 2025)

• Confidence-Aware Reasoning: Optimizing Self-Guided Thinking Trajectories in Large Reasoning Models (Zhang, EMNLP 2025)

ACL

• Jiaxin Zhang. 2025. Confidence-Aware Reasoning: Optimizing Self-Guided Thinking Trajectories in Large Reasoning Models. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 2081–2095, Suzhou (China). Association for Computational Linguistics.