<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="podolak-verma-2025-read">
    <titleInfo>
        <title>Read Your Own Mind: Reasoning Helps Surface Self-Confidence Signals in LLMs</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Jakub</namePart>
        <namePart type="family">Podolak</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Rajeev</namePart>
        <namePart type="family">Verma</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 2nd Workshop on Uncertainty-Aware NLP (UncertaiNLP 2025)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Noidea</namePart>
            <namePart type="family">Noidea</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">979-8-89176-349-4</identifier>
    </relatedItem>
    <abstract>We study the source of uncertainty in DeepSeek R1-32B by analyzing its self-reported verbal confidence on question answering (QA) tasks. In the default answer-then-confidence setting, the model is regularly over-confident, whereas semantic entropy - obtained by sampling many responses - remains reliable. We hypothesize that this is because of semantic entropy’s larger test-time compute, which lets us explore the model’s predictive distribution. We show that granting DeepSeek the budget to explore its distribution by forcing a long chain-of-thought before the final answer greatly improves its verbal score effectiveness, even on simple fact-retrieval questions that normally require no reasoning. Our analysis concludes that reliable uncertainty estimation requires explicit exploration of the generative space, and self-reported confidence is trustworthy only after such exploration.</abstract>
    <identifier type="citekey">podolak-verma-2025-read</identifier>
    <location>
        <url>https://aclanthology.org/2025.uncertainlp-main.21/</url>
    </location>
    <part>
        <date>2025-11</date>
        <extent unit="page">
            <start>247</start>
            <end>258</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Read Your Own Mind: Reasoning Helps Surface Self-Confidence Signals in LLMs
%A Podolak, Jakub
%A Verma, Rajeev
%Y Noidea, Noidea
%S Proceedings of the 2nd Workshop on Uncertainty-Aware NLP (UncertaiNLP 2025)
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-349-4
%F podolak-verma-2025-read
%X We study the source of uncertainty in DeepSeek R1-32B by analyzing its self-reported verbal confidence on question answering (QA) tasks. In the default answer-then-confidence setting, the model is regularly over-confident, whereas semantic entropy - obtained by sampling many responses - remains reliable. We hypothesize that this is because of semantic entropy’s larger test-time compute, which lets us explore the model’s predictive distribution. We show that granting DeepSeek the budget to explore its distribution by forcing a long chain-of-thought before the final answer greatly improves its verbal score effectiveness, even on simple fact-retrieval questions that normally require no reasoning. Our analysis concludes that reliable uncertainty estimation requires explicit exploration of the generative space, and self-reported confidence is trustworthy only after such exploration.
%U https://aclanthology.org/2025.uncertainlp-main.21/
%P 247-258

Markdown (Informal)

[Read Your Own Mind: Reasoning Helps Surface Self-Confidence Signals in LLMs](https://aclanthology.org/2025.uncertainlp-main.21/) (Podolak & Verma, UncertaiNLP 2025)

• Read Your Own Mind: Reasoning Helps Surface Self-Confidence Signals in LLMs (Podolak & Verma, UncertaiNLP 2025)

ACL

• Jakub Podolak and Rajeev Verma. 2025. Read Your Own Mind: Reasoning Helps Surface Self-Confidence Signals in LLMs. In Proceedings of the 2nd Workshop on Uncertainty-Aware NLP (UncertaiNLP 2025), pages 247–258, Suzhou, China. Association for Computational Linguistics.