<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="sadrieh-stefanik-2026-prompt">
    <titleInfo>
        <title>Prompt-Robust Language Models: Which Training Strategies Work?</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Frederic</namePart>
        <namePart type="family">Sadrieh</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Michal</namePart>
        <namePart type="family">Štefánik</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>Despite their strong performance, large language models remain highly sensitive to prompt formulation. Prior work has addressed this issue through refined data construction or more complex robustness-oriented training methods. In this work, we reproduce and compare these strategies under controlled conditions, evaluating their effects on worst-case prompt sensitivity and average performance. We find that conventional instruction fine-tuning offers limited reductions in prompt sensitivity: across methods, the average gap between the best- and worst-performing prompts remains 40–57% of performance. Moreover, recent robustness-enhancing methods often fail to outperform simpler data construction strategies; in worst-case settings, at least one simple method performs better in 5 of 8 evaluations. Overall, our findings show that prompt sensitivity remains an open problem, and motivate practitioners to prioritize simpler data construction techniques before adopting more sophisticated robustness methods.</abstract>
    <identifier type="citekey">sadrieh-stefanik-2026-prompt</identifier>
    <identifier type="doi">10.18653/2026.findings-emnlp.1900</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.1900/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>40490</start>
            <end>40507</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Prompt-Robust Language Models: Which Training Strategies Work?
%A Sadrieh, Frederic
%A Štefánik, Michal
%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 sadrieh-stefanik-2026-prompt
%X Despite their strong performance, large language models remain highly sensitive to prompt formulation. Prior work has addressed this issue through refined data construction or more complex robustness-oriented training methods. In this work, we reproduce and compare these strategies under controlled conditions, evaluating their effects on worst-case prompt sensitivity and average performance. We find that conventional instruction fine-tuning offers limited reductions in prompt sensitivity: across methods, the average gap between the best- and worst-performing prompts remains 40–57% of performance. Moreover, recent robustness-enhancing methods often fail to outperform simpler data construction strategies; in worst-case settings, at least one simple method performs better in 5 of 8 evaluations. Overall, our findings show that prompt sensitivity remains an open problem, and motivate practitioners to prioritize simpler data construction techniques before adopting more sophisticated robustness methods.
%R 10.18653/2026.findings-emnlp.1900
%U https://aclanthology.org/2026.findings-emnlp.1900/
%U https://doi.org/10.18653/2026.findings-emnlp.1900
%P 40490-40507

Markdown (Informal)

[Prompt-Robust Language Models: Which Training Strategies Work?](https://aclanthology.org/2026.findings-emnlp.1900/) (Sadrieh & Štefánik, Findings 2026)

• Prompt-Robust Language Models: Which Training Strategies Work? (Sadrieh & Štefánik, Findings 2026)

ACL

• Frederic Sadrieh and Michal Štefánik. 2026. Prompt-Robust Language Models: Which Training Strategies Work?. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 40490–40507, Budapest, Hungary. Association for Computational Linguistics.