<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="xu-etal-2026-steer">
    <titleInfo>
        <title>Steer on Rubrics: Auditable Steering of Schwartz Values in Language Models</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Zichen</namePart>
        <namePart type="family">Xu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Geng</namePart>
        <namePart type="family">Zhao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Aoxiang</namePart>
        <namePart type="family">Qin</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Li</namePart>
        <namePart type="family">Zhou</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Benyou</namePart>
        <namePart type="family">Wang</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
    </relatedItem>
    <abstract>Controlling the basic human values expressed by large language models (LLMs) has attracted growing attention in the mechanistic interpretability community. Recent work has made substantial attempts to steer such values through vector arithmetic or causal attribution over internal states. However, because values are complex social constructs, their mechanistic controllability still faces a major unresolved challenge: the auditability bottleneck, which makes it difficult to construct natural-language corpora that precisely capture value realization for mechanistic analysis. To address this challenge, we propose VALUEAUDITOR, a measurement-first framework that makes values behaviorally targetable before they are representationally steered. It operationalizes each value through grouped, theory-grounded response rubrics, derives graded supervision from same-scenario response contrasts, and performs inference-time steering through token-level sparse features that retain local cues of value expression. Experimental results show that VALUEAUDITOR significantly outperforms cutting-edge baseline methods on value steering benchmarks, while demonstrating robust control across values and scenarios with minimal trade-offs in response clarity. The code is publicly available at https://github.com/ZichenX/Steer-On-Rubrics</abstract>
    <identifier type="citekey">xu-etal-2026-steer</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.2147/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>48807</start>
            <end>48838</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Steer on Rubrics: Auditable Steering of Schwartz Values in Language Models
%A Xu, Zichen
%A Zhao, Geng
%A Qin, Aoxiang
%A Zhou, Li
%A Wang, Benyou
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F xu-etal-2026-steer
%X Controlling the basic human values expressed by large language models (LLMs) has attracted growing attention in the mechanistic interpretability community. Recent work has made substantial attempts to steer such values through vector arithmetic or causal attribution over internal states. However, because values are complex social constructs, their mechanistic controllability still faces a major unresolved challenge: the auditability bottleneck, which makes it difficult to construct natural-language corpora that precisely capture value realization for mechanistic analysis. To address this challenge, we propose VALUEAUDITOR, a measurement-first framework that makes values behaviorally targetable before they are representationally steered. It operationalizes each value through grouped, theory-grounded response rubrics, derives graded supervision from same-scenario response contrasts, and performs inference-time steering through token-level sparse features that retain local cues of value expression. Experimental results show that VALUEAUDITOR significantly outperforms cutting-edge baseline methods on value steering benchmarks, while demonstrating robust control across values and scenarios with minimal trade-offs in response clarity. The code is publicly available at https://github.com/ZichenX/Steer-On-Rubrics
%U https://aclanthology.org/2026.emnlp-main.2147/
%P 48807-48838

Markdown (Informal)

[Steer on Rubrics: Auditable Steering of Schwartz Values in Language Models](https://aclanthology.org/2026.emnlp-main.2147/) (Xu et al., EMNLP 2026)

• Steer on Rubrics: Auditable Steering of Schwartz Values in Language Models (Xu et al., EMNLP 2026)

ACL

• Zichen Xu, Geng Zhao, Aoxiang Qin, Li Zhou, and Benyou Wang. 2026. Steer on Rubrics: Auditable Steering of Schwartz Values in Language Models. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 48807–48838, Budapest, Hungary. Association for Computational Linguistics.