InstructionCP: A fast approach to transfer Large Language Models into target language

TL;DR

InstructionCP introduces an efficient method to adapt large language models to new languages by integrating instruction tags during continual pre-training, preserving conversational abilities and reducing resource needs.

Contribution

It proposes Instruction Continual Pre-training (InsCP), a novel technique that maintains conversational skills while adapting models to new languages using minimal data.

Findings

InsCP retains conversational and RLHF abilities.

It achieves effective language adaptation with only 0.1 billion tokens.

Experimental results confirm improved language alignment and reliability.

Abstract

The rapid development of large language models (LLMs) in recent years has largely focused on English, resulting in models that respond exclusively in English. To adapt these models to other languages, continual pre-training (CP) is often employed, followed by supervised fine-tuning (SFT) to maintain conversational abilities. However, CP and SFT can reduce a model's ability to filter harmful content. We propose Instruction Continual Pre-training (InsCP), which integrates instruction tags into the CP process to prevent loss of conversational proficiency while acquiring new languages. Our experiments demonstrate that InsCP retains conversational and Reinforcement Learning from Human Feedback (RLHF) abilities. Empirical evaluations on language alignment, reliability, and knowledge benchmarks confirm the efficacy of InsCP. Notably, this approach requires only 0.1 billion tokens of high-quality instruction-following data, thereby reducing resource consumption.