<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="aluguvelly-2026-one">
    <titleInfo>
        <title>One Vector, Thirty-Eight Languages: Benchmark-Derived Anti-Sycophancy Steering Works Worst Where Sycophancy Is Worst</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Avinash</namePart>
        <namePart type="given">Goutham</namePart>
        <namePart type="family">Aluguvelly</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>The Proceedings of the Workshop on Open Reasoning Across Cultures Languages (ORACLE)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Leonardo</namePart>
            <namePart type="family">Ranaldi</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Giulia</namePart>
            <namePart type="family">Pucci</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Federico</namePart>
            <namePart type="family">Ranaldi</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Marco</namePart>
            <namePart type="family">Valentino</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Alexandra</namePart>
            <namePart type="family">Birch</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Fabio</namePart>
            <namePart type="given">Massimo</namePart>
            <namePart type="family">Zanzotto</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">979-8-89176-483-5</identifier>
    </relatedItem>
    <abstract>Sycophancy, a model abandoning a correct position under social pressure, is worst in languages with thin instruction-tuning coverage, yet interventions are built almost entirely in English. We extract an anti-sycophancy steering direction from English data alone and apply it unchanged across 38 languages of a parallel benchmark, on two model families. Mean transfer falls monotonically across resource tiers: the vector keeps 48 percent of its English effect in high-resource languages, 19 percent in low-resource, and 3 percent in zeroshot ones. Baseline sycophancy is worst in that bottom tier, so the intervention is least available where the problem is largest. Neither redosing to a language’s own activation norms nor re-extracting the direction from its own pairs closes the gap, and the strength the benchmark criterion would have chosen steepens it. A specificity probe shows the direction does not isolate factual resistance: at its frozen setting, factual correction falls from 72.3 to 52.0 percent on the full English probe. A second direction, built from a contrast that flatly corrects the user, improves factual correction instead and does survive translation, in four of five translated high-resource languages and in Bengali once a general shift toward disagreement is subtracted. The steepest failure belongs to the benchmark’s operationalisation rather than to factual-resistance steering in general. A reduced check at 70B finds the same gradient.</abstract>
    <identifier type="citekey">aluguvelly-2026-one</identifier>
    <location>
        <url>https://aclanthology.org/2026.oracle-main.12/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>177</start>
            <end>193</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T One Vector, Thirty-Eight Languages: Benchmark-Derived Anti-Sycophancy Steering Works Worst Where Sycophancy Is Worst
%A Aluguvelly, Avinash Goutham
%Y Ranaldi, Leonardo
%Y Pucci, Giulia
%Y Ranaldi, Federico
%Y Valentino, Marco
%Y Birch, Alexandra
%Y Zanzotto, Fabio Massimo
%S The Proceedings of the Workshop on Open Reasoning Across Cultures Languages (ORACLE)
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-483-5
%F aluguvelly-2026-one
%X Sycophancy, a model abandoning a correct position under social pressure, is worst in languages with thin instruction-tuning coverage, yet interventions are built almost entirely in English. We extract an anti-sycophancy steering direction from English data alone and apply it unchanged across 38 languages of a parallel benchmark, on two model families. Mean transfer falls monotonically across resource tiers: the vector keeps 48 percent of its English effect in high-resource languages, 19 percent in low-resource, and 3 percent in zeroshot ones. Baseline sycophancy is worst in that bottom tier, so the intervention is least available where the problem is largest. Neither redosing to a language’s own activation norms nor re-extracting the direction from its own pairs closes the gap, and the strength the benchmark criterion would have chosen steepens it. A specificity probe shows the direction does not isolate factual resistance: at its frozen setting, factual correction falls from 72.3 to 52.0 percent on the full English probe. A second direction, built from a contrast that flatly corrects the user, improves factual correction instead and does survive translation, in four of five translated high-resource languages and in Bengali once a general shift toward disagreement is subtracted. The steepest failure belongs to the benchmark’s operationalisation rather than to factual-resistance steering in general. A reduced check at 70B finds the same gradient.
%U https://aclanthology.org/2026.oracle-main.12/
%P 177-193

Markdown (Informal)

[One Vector, Thirty-Eight Languages: Benchmark-Derived Anti-Sycophancy Steering Works Worst Where Sycophancy Is Worst](https://aclanthology.org/2026.oracle-main.12/) (Aluguvelly, ORACLE 2026)

• One Vector, Thirty-Eight Languages: Benchmark-Derived Anti-Sycophancy Steering Works Worst Where Sycophancy Is Worst (Aluguvelly, ORACLE 2026)

ACL

• Avinash Goutham Aluguvelly. 2026. One Vector, Thirty-Eight Languages: Benchmark-Derived Anti-Sycophancy Steering Works Worst Where Sycophancy Is Worst. In The Proceedings of the Workshop on Open Reasoning Across Cultures Languages (ORACLE), pages 177–193, Budapest, Hungary. Association for Computational Linguistics.