<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huang-etal-2026-llm-decide">
    <titleInfo>
        <title>Where Does an LLM Decide What’s Right? Sparse Normative Circuits Across Languages and Model Families</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Zhitao</namePart>
        <namePart type="family">Huang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Qiang</namePart>
        <namePart type="family">Huang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Wei</namePart>
        <namePart type="family">Zhai</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Lemao</namePart>
        <namePart type="family">Liu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yingmao</namePart>
        <namePart type="family">Tang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Dejing</namePart>
        <namePart type="family">Dou</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>Where does an LLM decide what’s right? We study whether normative judgment is implemented by a sparse, identifiable substrate or distributed across the model. Conventional circuit analysis can conflate normative computation with the multiple-choice format machinery that surrounds it. We pair four normative datasets (legal and moral, Chinese and English) with a non-normative MMLU control and apply a locate–disentangle–intervene pipeline that sets aside format-processing heads. Across Qwen3-8B, Llama3.1-8B, and Mistral-7B-v0.1, this isolates 5–10 normative-specific (NS) heads per model (under 1% of total) in MCQA settings. Targeted knockout and amplification reveal bidirectional causal effects across legal and moral tasks. The resulting heads form a sparse, late-layer intervention substrate for normative MCQA. Their coordinates differ across model families, so cross-model agreement is structural and circuit findings require functional re-identification in each family.</abstract>
    <identifier type="citekey">huang-etal-2026-llm-decide</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.1330/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>30329</start>
            <end>30351</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Where Does an LLM Decide What’s Right? Sparse Normative Circuits Across Languages and Model Families
%A Huang, Zhitao
%A Huang, Qiang
%A Zhai, Wei
%A Liu, Lemao
%A Tang, Yingmao
%A Dou, Dejing
%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 huang-etal-2026-llm-decide
%X Where does an LLM decide what’s right? We study whether normative judgment is implemented by a sparse, identifiable substrate or distributed across the model. Conventional circuit analysis can conflate normative computation with the multiple-choice format machinery that surrounds it. We pair four normative datasets (legal and moral, Chinese and English) with a non-normative MMLU control and apply a locate–disentangle–intervene pipeline that sets aside format-processing heads. Across Qwen3-8B, Llama3.1-8B, and Mistral-7B-v0.1, this isolates 5–10 normative-specific (NS) heads per model (under 1% of total) in MCQA settings. Targeted knockout and amplification reveal bidirectional causal effects across legal and moral tasks. The resulting heads form a sparse, late-layer intervention substrate for normative MCQA. Their coordinates differ across model families, so cross-model agreement is structural and circuit findings require functional re-identification in each family.
%U https://aclanthology.org/2026.emnlp-main.1330/
%P 30329-30351

Markdown (Informal)

[Where Does an LLM Decide What’s Right? Sparse Normative Circuits Across Languages and Model Families](https://aclanthology.org/2026.emnlp-main.1330/) (Huang et al., EMNLP 2026)

• Where Does an LLM Decide What’s Right? Sparse Normative Circuits Across Languages and Model Families (Huang et al., EMNLP 2026)

ACL

• Zhitao Huang, Qiang Huang, Wei Zhai, Lemao Liu, Yingmao Tang, and Dejing Dou. 2026. Where Does an LLM Decide What’s Right? Sparse Normative Circuits Across Languages and Model Families. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 30329–30351, Budapest, Hungary. Association for Computational Linguistics.