Salvato in:
Dettagli Bibliografici
Autori principali: Qu, Zhi, Wang, Yiran, Mao, Jiannan, Ding, Chenchen, Tanaka, Hideki, Utiyama, Masao, Watanabe, Taro
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2501.02979
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908379737227264
author Qu, Zhi
Wang, Yiran
Mao, Jiannan
Ding, Chenchen
Tanaka, Hideki
Utiyama, Masao
Watanabe, Taro
author_facet Qu, Zhi
Wang, Yiran
Mao, Jiannan
Ding, Chenchen
Tanaka, Hideki
Utiyama, Masao
Watanabe, Taro
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2501_02979
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Registering Source Tokens to Target Language Spaces in Multilingual Neural Machine Translation
Qu, Zhi
Wang, Yiran
Mao, Jiannan
Ding, Chenchen
Tanaka, Hideki
Utiyama, Masao
Watanabe, Taro
Computation and Language
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.
title Registering Source Tokens to Target Language Spaces in Multilingual Neural Machine Translation
topic Computation and Language
url https://arxiv.org/abs/2501.02979