FuxiMT: Sparsifying Large Language Models for Chinese-Centric Multilingual Machine Translation
This work addresses machine translation challenges for Chinese and 65 other languages, particularly in low-resource settings, representing an incremental improvement through model sparsification and curriculum learning.
The paper tackles multilingual machine translation with a focus on Chinese by developing FuxiMT, a sparsified large language model, which significantly outperforms state-of-the-art baselines, especially in low-resource scenarios, and shows strong zero-shot translation capabilities for unseen language pairs.
In this paper, we present FuxiMT, a novel Chinese-centric multilingual machine translation model powered by a sparsified large language model (LLM). We adopt a two-stage strategy to train FuxiMT. We first pre-train the model on a massive Chinese corpus and then conduct multilingual fine-tuning on a large parallel dataset encompassing 65 languages. FuxiMT incorporates Mixture-of-Experts (MoEs) and employs a curriculum learning strategy for robust performance across various resource levels. Experimental results demonstrate that FuxiMT significantly outperforms strong baselines, including state-of-the-art LLMs and machine translation models, particularly under low-resource scenarios. Furthermore, FuxiMT exhibits remarkable zero-shot translation capabilities for unseen language pairs, indicating its potential to bridge communication gaps where parallel data are scarce or unavailable.