TensorX
返回文献探索

Paper · arXiv 2407.05975

LLaMAX: Scaling Linguistic Horizons of LLM by Enhancing Translation Capabilities Beyond 100 Languages

Yinquan Lu, Wenhao Zhu, Lei Li, Yu Qiao, Fei Yuan

36 upvotesJuly 8, 2024arXiv 预印本
AI 摘要

LLaMAX enhances multilingual translation by extensive continual pre-training, achieving superior performance in 100+ languages without sacrificing generalization.

Large Language ModelsLLama seriesmultilingual continual pre-trainingvocabulary expansiondata augmentationspBLEUFlores-101 benchmarkM2M-100-12Bmultilingual foundation model

Abstract

Large Language Models~(LLMs) demonstrate remarkable translation capabilities in high-resource language tasks, yet their performance in low-resource languages is hindered by insufficient multilingual data during pre-training. To address this, we dedicate 35,000 A100-SXM4-80GB GPU hours in conducting extensive multilingual continual pre-training on the LLaMA series models, enabling translation support across more than 100 languages. Through a comprehensive analysis of training strategies, such as vocabulary expansion and data augmentation, we develop LLaMAX. Remarkably, without sacrificing its generalization ability, LLaMAX achieves significantly higher translation performance compared to existing open-source LLMs~(by more than 10 spBLEU points) and performs on-par with specialized translation model~(M2M-100-12B) on the Flores-101 benchmark. Extensive experiments indicate that LLaMAX can serve as a robust multilingual foundation model. The code~\url{https://github.com/CONE-MT/LLaMAX/.} and models~\url{https://huggingface.co/LLaMAX/.} are publicly available.

北京市昌平区探索星信息技术及软件开发工作室

京ICP备2026059466号
LLaMAX: Scaling Linguistic Horizons of LLM by Enhancing Translation Capabilities Beyond 100 Languages | TensorX