<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lu-etal-2026-hidden">
    <titleInfo>
        <title>Hidden Thoughts Are Not Secret: Reasoning-Trace Exposure in LLMs</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Yu-An</namePart>
        <namePart type="family">Lu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Ci-Yang</namePart>
        <namePart type="family">Tsai</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yu-Lin</namePart>
        <namePart type="family">Tsai</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Raluca</namePart>
        <namePart type="given">Ada</namePart>
        <namePart type="family">Popa</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Chia-Mu</namePart>
        <namePart type="family">Yu</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>Reasoning traces have become a valuable form of learning signals for improving and transferring the capabilities of large language models. In particular, detailed traces can help distill reasoning behavior from stronger teacher models into weaker student models. The value of capability transfer has motivated many deployed systems with reasoning models to hide raw internal traces and expose at most summaries and answers to users. As a result, we ask whether such interface-level trace hiding prevents users from obtaining useful reasoning supervision through prompting. We study this question with Reasoning Exposure Prompting (), a lightweight in-context elicitation method that uses shadow-model-generated demonstrations wrapped in auxiliary code-like formats to raise user-visible reasoning traces from a victim model. Across the common reasoning dataset, different victim models, and different student model distillation, substantially increases similarity between exposed and REP-conditioned internal traces while preserving useful reasoning signals.</abstract>
    <identifier type="citekey">lu-etal-2026-hidden</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.2303/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>49375</start>
            <end>49393</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Hidden Thoughts Are Not Secret: Reasoning-Trace Exposure in LLMs
%A Lu, Yu-An
%A Tsai, Ci-Yang
%A Tsai, Yu-Lin
%A Popa, Raluca Ada
%A Yu, Chia-Mu
%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 lu-etal-2026-hidden
%X Reasoning traces have become a valuable form of learning signals for improving and transferring the capabilities of large language models. In particular, detailed traces can help distill reasoning behavior from stronger teacher models into weaker student models. The value of capability transfer has motivated many deployed systems with reasoning models to hide raw internal traces and expose at most summaries and answers to users. As a result, we ask whether such interface-level trace hiding prevents users from obtaining useful reasoning supervision through prompting. We study this question with Reasoning Exposure Prompting (), a lightweight in-context elicitation method that uses shadow-model-generated demonstrations wrapped in auxiliary code-like formats to raise user-visible reasoning traces from a victim model. Across the common reasoning dataset, different victim models, and different student model distillation, substantially increases similarity between exposed and REP-conditioned internal traces while preserving useful reasoning signals.
%U https://aclanthology.org/2026.findings-emnlp.2303/
%P 49375-49393

Markdown (Informal)

[Hidden Thoughts Are Not Secret: Reasoning-Trace Exposure in LLMs](https://aclanthology.org/2026.findings-emnlp.2303/) (Lu et al., Findings 2026)

• Hidden Thoughts Are Not Secret: Reasoning-Trace Exposure in LLMs (Lu et al., Findings 2026)

ACL

• Yu-An Lu, Ci-Yang Tsai, Yu-Lin Tsai, Raluca Ada Popa, and Chia-Mu Yu. 2026. Hidden Thoughts Are Not Secret: Reasoning-Trace Exposure in LLMs. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 49375–49393, Budapest, Hungary. Association for Computational Linguistics.