@inproceedings{dinh-etal-2026-transporting,
title = "Transporting Concepts Across Languages: Optimal Transport Toward {E}nglish-Level Performance in Multilingual {LLM}s",
author = "Dinh, Anh Trac Duc and
Le, Anh Duc and
Quan, Tho",
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.861/",
pages = "19923--19943",
ISBN = "979-8-89176-459-0",
abstract = "Large language models (LLMs) exhibit strong English performance due to its dominance in pretraining data, yet remain inconsistent across non-English languages. Existing cross-lingual alignment methods rely on contrastive objectives that symmetrically perturb English representations, inadvertently degrading the proficiency they use as a reference. We propose an Optimal Transport (OT) framework that maps non-English hidden-state distributions toward English at middle transformer layers via entropy-regularised Sinkhorn iterations, while an English consistency regularisation term keeps the alignment anchor stationary. This asymmetric formulation preserves English performance by construction. Experiments on XNLI, XQuAD, and MMMLU show that our method delivers strong cross-lingual transfer while successfully avoiding the English performance degradation observed in the contrastive baseline."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="dinh-etal-2026-transporting">
<titleInfo>
<title>Transporting Concepts Across Languages: Optimal Transport Toward English-Level Performance in Multilingual LLMs</title>
</titleInfo>
<name type="personal">
<namePart type="given">Anh</namePart>
<namePart type="given">Trac</namePart>
<namePart type="given">Duc</namePart>
<namePart type="family">Dinh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anh</namePart>
<namePart type="given">Duc</namePart>
<namePart type="family">Le</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tho</namePart>
<namePart type="family">Quan</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>Large language models (LLMs) exhibit strong English performance due to its dominance in pretraining data, yet remain inconsistent across non-English languages. Existing cross-lingual alignment methods rely on contrastive objectives that symmetrically perturb English representations, inadvertently degrading the proficiency they use as a reference. We propose an Optimal Transport (OT) framework that maps non-English hidden-state distributions toward English at middle transformer layers via entropy-regularised Sinkhorn iterations, while an English consistency regularisation term keeps the alignment anchor stationary. This asymmetric formulation preserves English performance by construction. Experiments on XNLI, XQuAD, and MMMLU show that our method delivers strong cross-lingual transfer while successfully avoiding the English performance degradation observed in the contrastive baseline.</abstract>
<identifier type="citekey">dinh-etal-2026-transporting</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.861/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>19923</start>
<end>19943</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Transporting Concepts Across Languages: Optimal Transport Toward English-Level Performance in Multilingual LLMs
%A Dinh, Anh Trac Duc
%A Le, Anh Duc
%A Quan, Tho
%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 dinh-etal-2026-transporting
%X Large language models (LLMs) exhibit strong English performance due to its dominance in pretraining data, yet remain inconsistent across non-English languages. Existing cross-lingual alignment methods rely on contrastive objectives that symmetrically perturb English representations, inadvertently degrading the proficiency they use as a reference. We propose an Optimal Transport (OT) framework that maps non-English hidden-state distributions toward English at middle transformer layers via entropy-regularised Sinkhorn iterations, while an English consistency regularisation term keeps the alignment anchor stationary. This asymmetric formulation preserves English performance by construction. Experiments on XNLI, XQuAD, and MMMLU show that our method delivers strong cross-lingual transfer while successfully avoiding the English performance degradation observed in the contrastive baseline.
%U https://aclanthology.org/2026.emnlp-main.861/
%P 19923-19943
Markdown (Informal)
[Transporting Concepts Across Languages: Optimal Transport Toward English-Level Performance in Multilingual LLMs](https://aclanthology.org/2026.emnlp-main.861/) (Dinh et al., EMNLP 2026)
ACL