@inproceedings{zhang-2026-silent-locale,
title = "Silent Locale Corruption: Diagnosing and Mitigating {C}hinese-Locale Semantic Failures in {LLM} Tool Calling",
author = "Zhang, Ruiyang",
editor = "Kamalloo, Ehsan and
Gontier, Nicolas and
Rozanov, Nikolai and
Hashemzadeh, Maryam and
Arora, Akhil",
booktitle = "Proceedings of the Second Workshop for {RE}search on Agent Language Models ({REALM} 2026)",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.realm-1.3/",
pages = "27--37",
ISBN = "979-8-89176-485-9",
abstract = "When a language-model agent fills tool-call arguments, the argument value is the action: 转一万二 ({''}transfer twelve thousand'') must become transfer(12000), and 两斤苹果 ({''}two jin of apples'') must become order(weight{\_}kg=1.0). We show that today{'}s LLMs silently corrupt such values when Chinese locale semantics must be converted into canonical arguments {---} producing schema-valid, confidently wrong actions that no validator catches. We introduce ZhToolLocale, a diagnostic benchmark with machine-computed gold values (lunar-calendar libraries, deterministic numeral and unit parsers), paraphrase triplets, and a pre-registered audit protocol; we report 95{\%} bootstrap confidence intervals and significance tests for every headline claim. Across six models from five families {---} including two Chinese-trained {---} we find: (1) lunar-calendar conversion degrades significantly in every model (all p {\ensuremath{\leq}} .031) and catastrophically in five of six (0{--}28{\%}, vs. {\textasciitilde}100{\%} on fixed-date controls); a pre-registered matched-pair control refuted our initial memorization hypothesis {---} explicit lunar dates are, if anything, easier than festival names {---} indicating genuine conversion difficulty rather than recall; (2) Chinese-centric training confers a sizable advantage at the large tier (deepseek-chat: 93.3{\%} overall, 81{\%} lunar {---} the strongest model tested) but does not rescue the 7B tier (qwen2.5: lunar 3{\%}); (3) paraphrase disagreement detects scattered errors (precision up to 0.92) but is structurally blind to consistent mis-conversion, which we quantify; (4) a deterministic converter restores 94.8{--}100{\%} accuracy with significant gains for three models (all p {\ensuremath{\leq}} .0012), but routing is tier-governed: frontier models route voluntarily (93{--}99{\%}), a mid-tier model improves only under an explicit mandate, and two 7{--}8B models cannot be made to route at all (gains non-significant) {---} implying that deployable-tier safety must be architectural rather than prompted. We release the benchmark as a generator whose golds are computed on demand, with a per-category deployment card."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-2026-silent-locale">
<titleInfo>
<title>Silent Locale Corruption: Diagnosing and Mitigating Chinese-Locale Semantic Failures in LLM Tool Calling</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ruiyang</namePart>
<namePart type="family">Zhang</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 Second Workshop for REsearch on Agent Language Models (REALM 2026)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ehsan</namePart>
<namePart type="family">Kamalloo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nicolas</namePart>
<namePart type="family">Gontier</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nikolai</namePart>
<namePart type="family">Rozanov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maryam</namePart>
<namePart type="family">Hashemzadeh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Akhil</namePart>
<namePart type="family">Arora</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">979-8-89176-485-9</identifier>
</relatedItem>
<abstract>When a language-model agent fills tool-call arguments, the argument value is the action: 转一万二 (”transfer twelve thousand”) must become transfer(12000), and 两斤苹果 (”two jin of apples”) must become order(weight_kg=1.0). We show that today’s LLMs silently corrupt such values when Chinese locale semantics must be converted into canonical arguments — producing schema-valid, confidently wrong actions that no validator catches. We introduce ZhToolLocale, a diagnostic benchmark with machine-computed gold values (lunar-calendar libraries, deterministic numeral and unit parsers), paraphrase triplets, and a pre-registered audit protocol; we report 95% bootstrap confidence intervals and significance tests for every headline claim. Across six models from five families — including two Chinese-trained — we find: (1) lunar-calendar conversion degrades significantly in every model (all p \ensuremathłeq .031) and catastrophically in five of six (0–28%, vs. ~100% on fixed-date controls); a pre-registered matched-pair control refuted our initial memorization hypothesis — explicit lunar dates are, if anything, easier than festival names — indicating genuine conversion difficulty rather than recall; (2) Chinese-centric training confers a sizable advantage at the large tier (deepseek-chat: 93.3% overall, 81% lunar — the strongest model tested) but does not rescue the 7B tier (qwen2.5: lunar 3%); (3) paraphrase disagreement detects scattered errors (precision up to 0.92) but is structurally blind to consistent mis-conversion, which we quantify; (4) a deterministic converter restores 94.8–100% accuracy with significant gains for three models (all p \ensuremathłeq .0012), but routing is tier-governed: frontier models route voluntarily (93–99%), a mid-tier model improves only under an explicit mandate, and two 7–8B models cannot be made to route at all (gains non-significant) — implying that deployable-tier safety must be architectural rather than prompted. We release the benchmark as a generator whose golds are computed on demand, with a per-category deployment card.</abstract>
<identifier type="citekey">zhang-2026-silent-locale</identifier>
<location>
<url>https://aclanthology.org/2026.realm-1.3/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>27</start>
<end>37</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Silent Locale Corruption: Diagnosing and Mitigating Chinese-Locale Semantic Failures in LLM Tool Calling
%A Zhang, Ruiyang
%Y Kamalloo, Ehsan
%Y Gontier, Nicolas
%Y Rozanov, Nikolai
%Y Hashemzadeh, Maryam
%Y Arora, Akhil
%S Proceedings of the Second Workshop for REsearch on Agent Language Models (REALM 2026)
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-485-9
%F zhang-2026-silent-locale
%X When a language-model agent fills tool-call arguments, the argument value is the action: 转一万二 (”transfer twelve thousand”) must become transfer(12000), and 两斤苹果 (”two jin of apples”) must become order(weight_kg=1.0). We show that today’s LLMs silently corrupt such values when Chinese locale semantics must be converted into canonical arguments — producing schema-valid, confidently wrong actions that no validator catches. We introduce ZhToolLocale, a diagnostic benchmark with machine-computed gold values (lunar-calendar libraries, deterministic numeral and unit parsers), paraphrase triplets, and a pre-registered audit protocol; we report 95% bootstrap confidence intervals and significance tests for every headline claim. Across six models from five families — including two Chinese-trained — we find: (1) lunar-calendar conversion degrades significantly in every model (all p \ensuremathłeq .031) and catastrophically in five of six (0–28%, vs. ~100% on fixed-date controls); a pre-registered matched-pair control refuted our initial memorization hypothesis — explicit lunar dates are, if anything, easier than festival names — indicating genuine conversion difficulty rather than recall; (2) Chinese-centric training confers a sizable advantage at the large tier (deepseek-chat: 93.3% overall, 81% lunar — the strongest model tested) but does not rescue the 7B tier (qwen2.5: lunar 3%); (3) paraphrase disagreement detects scattered errors (precision up to 0.92) but is structurally blind to consistent mis-conversion, which we quantify; (4) a deterministic converter restores 94.8–100% accuracy with significant gains for three models (all p \ensuremathłeq .0012), but routing is tier-governed: frontier models route voluntarily (93–99%), a mid-tier model improves only under an explicit mandate, and two 7–8B models cannot be made to route at all (gains non-significant) — implying that deployable-tier safety must be architectural rather than prompted. We release the benchmark as a generator whose golds are computed on demand, with a per-category deployment card.
%U https://aclanthology.org/2026.realm-1.3/
%P 27-37
Markdown (Informal)
[Silent Locale Corruption: Diagnosing and Mitigating Chinese-Locale Semantic Failures in LLM Tool Calling](https://aclanthology.org/2026.realm-1.3/) (Zhang, REALM 2026)
ACL