A Novel Paradigm Boosting Translation Capabilities of Large Language Models

arXiv:2403.11430v1 Announce Type: new
Abstract: This paper presents a study on strategies to enhance the translation capabilities of large language models (LLMs) in the context of machine translation (MT) tasks. The paper proposes a novel paradigm consisting of three stages: Secondary Pre-training using Extensive Monolingual Data, Continual Pre-training with Interlinear Text Format Documents, and Leveraging Source-Language Consistent Instruction for Supervised Fine-Tuning. Previous research on LLMs focused on various strategies for supervised fine-tuning (SFT), but their effectiveness has been limited. While …

abstract arxiv boosting capabilities context continual cs.cl data language language models large language large language models llms machine machine translation novel paper paradigm pre-training strategies study tasks training translation type