<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lu-etal-2026-knowing">
    <titleInfo>
        <title>Knowing When to Speak: Aligning LLM Honesty via Solvability-Aware Preference Optimization</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Wenpeng</namePart>
        <namePart type="family">Lu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Xueguan</namePart>
        <namePart type="family">Zhao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Shuxia</namePart>
        <namePart type="family">Lin</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Linhai</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Chaoqun</namePart>
        <namePart type="family">Zheng</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Weiyu</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jiasheng</namePart>
        <namePart type="family">Si</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 Language Models often exhibit dishonest behaviors–confidently hallucinating answers to questions beyond their knowledge boundary or hedging on queries they have mastered. Although existing mitigation methods have progressed to establishing the model knowledge boundary, this boundary is still typically treated as a coarse “known vs. unknown” distinction, leaving a finer-grained question underexplored: Does a failure reflect missing knowledge, or merely under-expressed capability? In this paper, we propose KCAPO, a framework that aligns the model’s output behavior with its latent capabilities. It estimates query-level solvability with a consistency-supervised hidden-state probe, then uses its mismatch with response confidence and factual behavior to refine the knowledge boundary into four capability–behavior cases. Based on this fine-grained characterization, we construct preference data to dynamically “Wake Up” dormant knowledge or “Suppress” blind guessing. Furthermore, we introduce an adaptive margin into the DPO objective, which prioritizes correcting responses with severe cognitive dissonance. Extensive experiments on short-form QA and refusal benchmarks show that KCAPO reduces hallucinations without compromising utility and improves selective abstention.</abstract>
    <identifier type="citekey">lu-etal-2026-knowing</identifier>
    <identifier type="doi">10.18653/2026.findings-emnlp.2451</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.2451/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>52609</start>
            <end>52626</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Knowing When to Speak: Aligning LLM Honesty via Solvability-Aware Preference Optimization
%A Lu, Wenpeng
%A Zhao, Xueguan
%A Lin, Shuxia
%A Zhang, Linhai
%A Zheng, Chaoqun
%A Zhang, Weiyu
%A Si, Jiasheng
%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-knowing
%X Large Language Models often exhibit dishonest behaviors–confidently hallucinating answers to questions beyond their knowledge boundary or hedging on queries they have mastered. Although existing mitigation methods have progressed to establishing the model knowledge boundary, this boundary is still typically treated as a coarse “known vs. unknown” distinction, leaving a finer-grained question underexplored: Does a failure reflect missing knowledge, or merely under-expressed capability? In this paper, we propose KCAPO, a framework that aligns the model’s output behavior with its latent capabilities. It estimates query-level solvability with a consistency-supervised hidden-state probe, then uses its mismatch with response confidence and factual behavior to refine the knowledge boundary into four capability–behavior cases. Based on this fine-grained characterization, we construct preference data to dynamically “Wake Up” dormant knowledge or “Suppress” blind guessing. Furthermore, we introduce an adaptive margin into the DPO objective, which prioritizes correcting responses with severe cognitive dissonance. Extensive experiments on short-form QA and refusal benchmarks show that KCAPO reduces hallucinations without compromising utility and improves selective abstention.
%R 10.18653/2026.findings-emnlp.2451
%U https://aclanthology.org/2026.findings-emnlp.2451/
%U https://doi.org/10.18653/2026.findings-emnlp.2451
%P 52609-52626

Markdown (Informal)

[Knowing When to Speak: Aligning LLM Honesty via Solvability-Aware Preference Optimization](https://aclanthology.org/2026.findings-emnlp.2451/) (Lu et al., Findings 2026)

• Knowing When to Speak: Aligning LLM Honesty via Solvability-Aware Preference Optimization (Lu et al., Findings 2026)

ACL

• Wenpeng Lu, Xueguan Zhao, Shuxia Lin, Linhai Zhang, Chaoqun Zheng, Weiyu Zhang, and Jiasheng Si. 2026. Knowing When to Speak: Aligning LLM Honesty via Solvability-Aware Preference Optimization. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 52609–52626, Budapest, Hungary. Association for Computational Linguistics.