Query-Following vs Context-Anchoring: How LLMs Handle Cross-Turn Language Switching

Kyuhee Kim; Chengheng Li Chen; Anna Sotnikova

Query-Following vs Context-Anchoring: How LLMs Handle Cross-Turn Language Switching

Kyuhee Kim, Chengheng Li Chen, Anna Sotnikova

Correct Metadata for

Use this form to create a GitHub issue with structured data describing the correction. You will need a GitHub account. Once you create that issue, the correction will be reviewed by a staff member.

⚠️ Mobile Users: Submitting this form to create a new issue will only work with github.com, not the GitHub Mobile app.

Important: The Anthology treat PDFs as authoritative. Please use this form only to correct data that is out of line with the PDF. See our corrections guidelines if you need to change the PDF.

Title Adjust the title. Retain tags such as <fixed-case>.

Authors Adjust author names and order to match the PDF.

Abstract Correct abstract if needed. Retain XML formatting tags such as <tex-math>. You may use <b>...</b> for bold, <i>...</i> for italic, and <url>...</url> for URLs.

Verification against PDF Ensure that the new title/authors match the snapshot below. (If there is no snapshot or it is too small, consult the PDF.)

Authors concatenated from the text boxes above:

ALL author names match the snapshot above—including middle initials, hyphens, and accents.

Abstract

When multilingual users switch languages mid-conversation, how should LLMs respond? We extend MultiChallenge to evaluate cross-turn language switching, translating 182 multi-turn conversations into German, Chinese, Spanish, and Arabic. Across five frontier models, we observe asymmetric behavior: switching into a foreign language (EN→X) yields high query-language fidelity (89–99%), but switching back to English (X→EN) reveals divergent policies. GPT-5 follows the query language (>95%), while Claude Opus 4.5 and Command R+ maintain the established conversation language (<8%). Task accuracy remains stable across conditions regardless of language selection differences. A simple explicit system prompt shows limited effectiveness in modifying these defaults.

Anthology ID:: 2026.mme-main.13
Volume:: Proceedings of the First Workshop on Multilingual Multicultural Evaluation
Month:: March
Year:: 2026
Address:: Rabat, Morocco
Editors:: Pinzhen Chen, Vilém Zouhar, Hanxu Hu, Simran Khanuja, Wenhao Zhu, Barry Haddow, Alexandra Birch, Alham Fikri Aji, Rico Sennrich, Sara Hooker
Venues:: MME | WS
SIG:
Publisher:: Association for Computational Linguistics
Note:
Pages:: 196–203
Language:
URL:: https://aclanthology.org/2026.mme-main.13/
DOI:
Bibkey:
Cite (ACL):: Kyuhee Kim, Chengheng Li Chen, and Anna Sotnikova. 2026. Query-Following vs Context-Anchoring: How LLMs Handle Cross-Turn Language Switching. In Proceedings of the First Workshop on Multilingual Multicultural Evaluation, pages 196–203, Rabat, Morocco. Association for Computational Linguistics.
Cite (Informal):: Query-Following vs Context-Anchoring: How LLMs Handle Cross-Turn Language Switching (Kim et al., MME 2026)
Copy Citation:
PDF:: https://aclanthology.org/2026.mme-main.13.pdf

PDF Cite Search Fix data

Export citation

BibTeX
MODS XML
Endnote
Preformatted

@inproceedings{kim-etal-2026-query,
    title = "Query-Following vs Context-Anchoring: How {LLM}s Handle Cross-Turn Language Switching",
    author = "Kim, Kyuhee  and
      Chen, Chengheng Li  and
      Sotnikova, Anna",
    editor = "Chen, Pinzhen  and
      Zouhar, Vil{\'e}m  and
      Hu, Hanxu  and
      Khanuja, Simran  and
      Zhu, Wenhao  and
      Haddow, Barry  and
      Birch, Alexandra  and
      Aji, Alham Fikri  and
      Sennrich, Rico  and
      Hooker, Sara",
    booktitle = "Proceedings of the First Workshop on Multilingual Multicultural Evaluation",
    month = mar,
    year = "2026",
    address = "Rabat, Morocco",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2026.mme-main.13/",
    pages = "196--203",
    ISBN = "979-8-89176-368-5",
    abstract = "When multilingual users switch languages mid-conversation, how should LLMs respond? We extend MultiChallenge to evaluate cross-turn language switching, translating 182 multi-turn conversations into German, Chinese, Spanish, and Arabic. Across five frontier models, we observe asymmetric behavior: switching into a foreign language (EN{\textrightarrow}X) yields high query-language fidelity (89{--}99{\%}), but switching back to English (X{\textrightarrow}EN) reveals divergent policies. GPT-5 follows the query language ($>$95{\%}), while Claude Opus 4.5 and Command R+ maintain the established conversation language ($<$8{\%}). Task accuracy remains stable across conditions regardless of language selection differences. A simple explicit system prompt shows limited effectiveness in modifying these defaults."
}

Download as File

<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kim-etal-2026-query">
    <titleInfo>
        <title>Query-Following vs Context-Anchoring: How LLMs Handle Cross-Turn Language Switching</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Kyuhee</namePart>
        <namePart type="family">Kim</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Chengheng</namePart>
        <namePart type="given">Li</namePart>
        <namePart type="family">Chen</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Anna</namePart>
        <namePart type="family">Sotnikova</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-03</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the First Workshop on Multilingual Multicultural Evaluation</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Pinzhen</namePart>
            <namePart type="family">Chen</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Vilém</namePart>
            <namePart type="family">Zouhar</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Hanxu</namePart>
            <namePart type="family">Hu</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Simran</namePart>
            <namePart type="family">Khanuja</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Wenhao</namePart>
            <namePart type="family">Zhu</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Barry</namePart>
            <namePart type="family">Haddow</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">Alham</namePart>
            <namePart type="given">Fikri</namePart>
            <namePart type="family">Aji</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Rico</namePart>
            <namePart type="family">Sennrich</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Sara</namePart>
            <namePart type="family">Hooker</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Rabat, Morocco</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-368-5</identifier>
    </relatedItem>
    <abstract>When multilingual users switch languages mid-conversation, how should LLMs respond? We extend MultiChallenge to evaluate cross-turn language switching, translating 182 multi-turn conversations into German, Chinese, Spanish, and Arabic. Across five frontier models, we observe asymmetric behavior: switching into a foreign language (EN→X) yields high query-language fidelity (89–99%), but switching back to English (X→EN) reveals divergent policies. GPT-5 follows the query language (&gt;95%), while Claude Opus 4.5 and Command R+ maintain the established conversation language (&lt;8%). Task accuracy remains stable across conditions regardless of language selection differences. A simple explicit system prompt shows limited effectiveness in modifying these defaults.</abstract>
    <identifier type="citekey">kim-etal-2026-query</identifier>
    <location>
        <url>https://aclanthology.org/2026.mme-main.13/</url>
    </location>
    <part>
        <date>2026-03</date>
        <extent unit="page">
            <start>196</start>
            <end>203</end>
        </extent>
    </part>
</mods>
</modsCollection>

Download as File

%0 Conference Proceedings
%T Query-Following vs Context-Anchoring: How LLMs Handle Cross-Turn Language Switching
%A Kim, Kyuhee
%A Chen, Chengheng Li
%A Sotnikova, Anna
%Y Chen, Pinzhen
%Y Zouhar, Vilém
%Y Hu, Hanxu
%Y Khanuja, Simran
%Y Zhu, Wenhao
%Y Haddow, Barry
%Y Birch, Alexandra
%Y Aji, Alham Fikri
%Y Sennrich, Rico
%Y Hooker, Sara
%S Proceedings of the First Workshop on Multilingual Multicultural Evaluation
%D 2026
%8 March
%I Association for Computational Linguistics
%C Rabat, Morocco
%@ 979-8-89176-368-5
%F kim-etal-2026-query
%X When multilingual users switch languages mid-conversation, how should LLMs respond? We extend MultiChallenge to evaluate cross-turn language switching, translating 182 multi-turn conversations into German, Chinese, Spanish, and Arabic. Across five frontier models, we observe asymmetric behavior: switching into a foreign language (EN→X) yields high query-language fidelity (89–99%), but switching back to English (X→EN) reveals divergent policies. GPT-5 follows the query language (>95%), while Claude Opus 4.5 and Command R+ maintain the established conversation language (<8%). Task accuracy remains stable across conditions regardless of language selection differences. A simple explicit system prompt shows limited effectiveness in modifying these defaults.
%U https://aclanthology.org/2026.mme-main.13/
%P 196-203

Download as File

Markdown (Informal)

[Query-Following vs Context-Anchoring: How LLMs Handle Cross-Turn Language Switching](https://aclanthology.org/2026.mme-main.13/) (Kim et al., MME 2026)

Query-Following vs Context-Anchoring: How LLMs Handle Cross-Turn Language Switching (Kim et al., MME 2026)

ACL

Kyuhee Kim, Chengheng Li Chen, and Anna Sotnikova. 2026. Query-Following vs Context-Anchoring: How LLMs Handle Cross-Turn Language Switching. In Proceedings of the First Workshop on Multilingual Multicultural Evaluation, pages 196–203, Rabat, Morocco. Association for Computational Linguistics.