@inproceedings{abdo-etal-2026-translation,
title = "Translation-Induced Label Drift in Natural Language Inference: Evidence from Nine Languages",
author = {Abdo, Muhammad S. and
Artkaew, Phakphum and
Ashraf, Sadia and
Paiva, Elder Henrique Attala E and
Benavides, Jose and
Karkoutli, Hale and
Lieffers, Melissa and
Naske, Elias and
Shi, Wenchen and
Wang, Jinyuan and
K{\"u}bler, Sandra},
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.817/",
pages = "19014--19027",
ISBN = "979-8-89176-459-0",
abstract = "Multilingual natural language inference (NLI) benchmarks are routinely constructed by translating English datasets into target languages, based on the assumption that inference relations between premise{--}hypothesis pairs are preserved under translation. This assumption, however, has rarely been tested empirically at scale. We investigate this by conducting a large-scale human annotation study across nine typologically diverse languages: Arabic, German, Spanish, Portuguese, Thai, Urdu, Turkish, Mandarin, and Cantonese. We translated 1,000 English premise{--}hypothesis pairs, sampled across four linguistically challenging phenomena, and annotated each target-language version by a native-speaker linguist. Additionally, a native English speaker annotated the source pairs under the same protocol, allowing us to distinguish source-label disagreement from translation-induced change. While agreement between translated-language annotations and the original dataset labels ranges from 55.1{\%} to 67.4{\%}, agreement with the native-English annotation ranges from 57.8{\%} to 64.2{\%}. Entailment and Contradiction frequently shift toward Neutral, but the English control shows that some apparent drift is already present in the source items. These results call into question using transfer of English labels and highlight the need for native-language annotation in multilingual NLI evaluation."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="abdo-etal-2026-translation">
<titleInfo>
<title>Translation-Induced Label Drift in Natural Language Inference: Evidence from Nine Languages</title>
</titleInfo>
<name type="personal">
<namePart type="given">Muhammad</namePart>
<namePart type="given">S</namePart>
<namePart type="family">Abdo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Phakphum</namePart>
<namePart type="family">Artkaew</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sadia</namePart>
<namePart type="family">Ashraf</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Elder</namePart>
<namePart type="given">Henrique</namePart>
<namePart type="given">Attala</namePart>
<namePart type="given">E</namePart>
<namePart type="family">Paiva</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jose</namePart>
<namePart type="family">Benavides</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hale</namePart>
<namePart type="family">Karkoutli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Melissa</namePart>
<namePart type="family">Lieffers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Elias</namePart>
<namePart type="family">Naske</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wenchen</namePart>
<namePart type="family">Shi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jinyuan</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sandra</namePart>
<namePart type="family">Kübler</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>Multilingual natural language inference (NLI) benchmarks are routinely constructed by translating English datasets into target languages, based on the assumption that inference relations between premise–hypothesis pairs are preserved under translation. This assumption, however, has rarely been tested empirically at scale. We investigate this by conducting a large-scale human annotation study across nine typologically diverse languages: Arabic, German, Spanish, Portuguese, Thai, Urdu, Turkish, Mandarin, and Cantonese. We translated 1,000 English premise–hypothesis pairs, sampled across four linguistically challenging phenomena, and annotated each target-language version by a native-speaker linguist. Additionally, a native English speaker annotated the source pairs under the same protocol, allowing us to distinguish source-label disagreement from translation-induced change. While agreement between translated-language annotations and the original dataset labels ranges from 55.1% to 67.4%, agreement with the native-English annotation ranges from 57.8% to 64.2%. Entailment and Contradiction frequently shift toward Neutral, but the English control shows that some apparent drift is already present in the source items. These results call into question using transfer of English labels and highlight the need for native-language annotation in multilingual NLI evaluation.</abstract>
<identifier type="citekey">abdo-etal-2026-translation</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.817/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>19014</start>
<end>19027</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Translation-Induced Label Drift in Natural Language Inference: Evidence from Nine Languages
%A Abdo, Muhammad S.
%A Artkaew, Phakphum
%A Ashraf, Sadia
%A Paiva, Elder Henrique Attala E.
%A Benavides, Jose
%A Karkoutli, Hale
%A Lieffers, Melissa
%A Naske, Elias
%A Shi, Wenchen
%A Wang, Jinyuan
%A Kübler, Sandra
%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 abdo-etal-2026-translation
%X Multilingual natural language inference (NLI) benchmarks are routinely constructed by translating English datasets into target languages, based on the assumption that inference relations between premise–hypothesis pairs are preserved under translation. This assumption, however, has rarely been tested empirically at scale. We investigate this by conducting a large-scale human annotation study across nine typologically diverse languages: Arabic, German, Spanish, Portuguese, Thai, Urdu, Turkish, Mandarin, and Cantonese. We translated 1,000 English premise–hypothesis pairs, sampled across four linguistically challenging phenomena, and annotated each target-language version by a native-speaker linguist. Additionally, a native English speaker annotated the source pairs under the same protocol, allowing us to distinguish source-label disagreement from translation-induced change. While agreement between translated-language annotations and the original dataset labels ranges from 55.1% to 67.4%, agreement with the native-English annotation ranges from 57.8% to 64.2%. Entailment and Contradiction frequently shift toward Neutral, but the English control shows that some apparent drift is already present in the source items. These results call into question using transfer of English labels and highlight the need for native-language annotation in multilingual NLI evaluation.
%U https://aclanthology.org/2026.emnlp-main.817/
%P 19014-19027
Markdown (Informal)
[Translation-Induced Label Drift in Natural Language Inference: Evidence from Nine Languages](https://aclanthology.org/2026.emnlp-main.817/) (Abdo et al., EMNLP 2026)
ACL
- Muhammad S. Abdo, Phakphum Artkaew, Sadia Ashraf, Elder Henrique Attala E Paiva, Jose Benavides, Hale Karkoutli, Melissa Lieffers, Elias Naske, Wenchen Shi, Jinyuan Wang, and Sandra Kübler. 2026. Translation-Induced Label Drift in Natural Language Inference: Evidence from Nine Languages. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 19014–19027, Budapest, Hungary. Association for Computational Linguistics.