aiqbench takeAI Research

This research investigates how scaling both model parameters and training data impacts the multilingual translation capabilities of open-source LLMs. By adapting the Gemma3 architecture through continual pretraining, the authors developed MiLMMT-46 to achieve state-of-the-art performance across 46 languages without relying on proprietary systems. The study provides critical insights into optimizing resource efficiency for diverse language coverage in next-generation translation models.

Source coveringarXiv

What this isA model's reading of the article and related coverage, written for aiqbench. Not the publisher's words.

Scaling Model and Data for Multilingual Machine Translation with Open Large Language Models

arXivJuly 21, 20262 min read

arXiv:2602.11961v3 Announce Type: replace Abstract: Open large language models (LLMs) have demonstrated improving multilingual capabilities in recent years. In this paper, we present a study of open LLMs for multilingual machine translation (MT) across a range of languages, and investigate the effects of model scaling and data scaling when adapting open LLMs to multilingual MT through continual pretraining and instruction finetuning. Based on the Gemma3 model family, we develop MiLMMT-46, which achieves top-tier multilingual translation performance across 46 languages. Extensive experiments show that MiLMMT-46 consistently outperforms recent state-of-the-art (SOTA) models, including Seed-X, HY-MT-1.5, and TranslateGemma, and achieves competitive performance with strong proprietary systems such as Google Translate and Gemini 3 Pro. Models are released at https://huggingface.co/collections/xiaomi-research/milmmt-46. Codes are released at https://github.com/xiaomi-research/gemmax.

Abstract:Open large language models (LLMs) have demonstrated improving multilingual capabilities in recent years. In this paper, we present a study of open LLMs for multilingual machine translation (MT) across a range of languages, and investigate the effects of model scaling and data scaling when adapting open LLMs to multilingual MT through continual pretraining and instruction finetuning. Based on the Gemma3 model family, we develop MiLMMT-46, which achieves top-tier multilingual translation performance across 46 languages. Extensive experiments show that MiLMMT-46 consistently outperforms recent state-of-the-art (SOTA) models, including Seed-X, HY-MT-1.5, and TranslateGemma, and achieves competitive performance with strong proprietary systems such as Google Translate and Gemini 3 Pro. Models are released at this https URL. Codes are released at this https URL.

Computation and Language (cs.CL)

arXiv:2602.11961 [cs.CL]

(or arXiv:2602.11961v3 [cs.CL] for this version)

Full text under the publisher's open licence · 132 words

Read at arXiv ↗Ask about this article