@inproceedings{qu-etal-2025-registering,
title = "Registering Source Tokens to Target Language Spaces in Multilingual Neural Machine Translation",
author = "Qu, Zhi and
Wang, Yiran and
Mao, Jiannan and
Ding, Chenchen and
Tanaka, Hideki and
Utiyama, Masao and
Watanabe, Taro",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.acl-long.1052/",
doi = "10.18653/v1/2025.acl-long.1052",
pages = "21687--21706",
ISBN = "979-8-89176-251-0",
abstract = "The multilingual neural machine translation (MNMT) aims for arbitrary translations across multiple languages.Although MNMT-specific models trained on parallel data offer low costs in training and deployment, their performance consistently lags behind that of large language models (LLMs).In this work, we introduce registering, a novel method that enables a small MNMT-specific model to compete with LLMs.Specifically, we insert a set of artificial tokens specifying the target language, called registers, into the input sequence between the source and target tokens.By modifying the attention mask, the target token generation only pays attention to the activation of registers, representing the source tokens in the target language space.Experiments on EC-40, a large-scale benchmark, show that our method advances the state-of-the-art of MNMT.We further pre-train two models, namely MITRE (multilingual translation with registers), by 9.3 billion sentence pairs across 24 languages collected from public corpora.One of them, MITRE-913M, outperforms NLLB-3.3B, achieves comparable performance with commercial LLMs, and shows strong adaptability in fine-tuning.Finally, we open-source our models to facilitate further research and development in MNMT: https://github.com/zhiqu22/mitre."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="qu-etal-2025-registering">
<titleInfo>
<title>Registering Source Tokens to Target Language Spaces in Multilingual Neural Machine Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zhi</namePart>
<namePart type="family">Qu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yiran</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiannan</namePart>
<namePart type="family">Mao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chenchen</namePart>
<namePart type="family">Ding</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hideki</namePart>
<namePart type="family">Tanaka</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Masao</namePart>
<namePart type="family">Utiyama</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Taro</namePart>
<namePart type="family">Watanabe</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-251-0</identifier>
</relatedItem>
<abstract>The multilingual neural machine translation (MNMT) aims for arbitrary translations across multiple languages.Although MNMT-specific models trained on parallel data offer low costs in training and deployment, their performance consistently lags behind that of large language models (LLMs).In this work, we introduce registering, a novel method that enables a small MNMT-specific model to compete with LLMs.Specifically, we insert a set of artificial tokens specifying the target language, called registers, into the input sequence between the source and target tokens.By modifying the attention mask, the target token generation only pays attention to the activation of registers, representing the source tokens in the target language space.Experiments on EC-40, a large-scale benchmark, show that our method advances the state-of-the-art of MNMT.We further pre-train two models, namely MITRE (multilingual translation with registers), by 9.3 billion sentence pairs across 24 languages collected from public corpora.One of them, MITRE-913M, outperforms NLLB-3.3B, achieves comparable performance with commercial LLMs, and shows strong adaptability in fine-tuning.Finally, we open-source our models to facilitate further research and development in MNMT: https://github.com/zhiqu22/mitre.</abstract>
<identifier type="citekey">qu-etal-2025-registering</identifier>
<identifier type="doi">10.18653/v1/2025.acl-long.1052</identifier>
<location>
<url>https://aclanthology.org/2025.acl-long.1052/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>21687</start>
<end>21706</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Registering Source Tokens to Target Language Spaces in Multilingual Neural Machine Translation
%A Qu, Zhi
%A Wang, Yiran
%A Mao, Jiannan
%A Ding, Chenchen
%A Tanaka, Hideki
%A Utiyama, Masao
%A Watanabe, Taro
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-251-0
%F qu-etal-2025-registering
%X The multilingual neural machine translation (MNMT) aims for arbitrary translations across multiple languages.Although MNMT-specific models trained on parallel data offer low costs in training and deployment, their performance consistently lags behind that of large language models (LLMs).In this work, we introduce registering, a novel method that enables a small MNMT-specific model to compete with LLMs.Specifically, we insert a set of artificial tokens specifying the target language, called registers, into the input sequence between the source and target tokens.By modifying the attention mask, the target token generation only pays attention to the activation of registers, representing the source tokens in the target language space.Experiments on EC-40, a large-scale benchmark, show that our method advances the state-of-the-art of MNMT.We further pre-train two models, namely MITRE (multilingual translation with registers), by 9.3 billion sentence pairs across 24 languages collected from public corpora.One of them, MITRE-913M, outperforms NLLB-3.3B, achieves comparable performance with commercial LLMs, and shows strong adaptability in fine-tuning.Finally, we open-source our models to facilitate further research and development in MNMT: https://github.com/zhiqu22/mitre.
%R 10.18653/v1/2025.acl-long.1052
%U https://aclanthology.org/2025.acl-long.1052/
%U https://doi.org/10.18653/v1/2025.acl-long.1052
%P 21687-21706
Markdown (Informal)
[Registering Source Tokens to Target Language Spaces in Multilingual Neural Machine Translation](https://aclanthology.org/2025.acl-long.1052/) (Qu et al., ACL 2025)
ACL