Monolingual or Multilingual Instruction Tuning: Which Makes a Better Alpaca

Anonymous

Abstract: Foundation large language models (LLMs) can be instruction-tuned to perform open-domain question answering, facilitating applications such as AI assistants. While such efforts are often carried out in a single language, we empirically analyze cost-efficient approaches to multilingual tuning. Our study employs the Alpaca dataset and machine translations of it to form multilingual training data, which is then used to tune LLMs through low-rank adaptation and full-parameter training. Under a fixed budget, comparisons show that multilingual tuning is on par or better than separately tuning a model for each language. Further, multilingual tuning with downsampled data can be as powerful and more robust. Our findings serve as a guide for expanding language support through instruction tuning with constrained computational resources.

Paper Type: short

Research Area: NLP Applications

Contribution Types: NLP engineering experiment, Approaches low compute settings-efficiency

Languages Studied: English, Spanish, French, Bulgarian, Russian, Chinese, Bengali, Norwegian

Consent To Share Submission Details: On behalf of all authors, we agree to the terms above to share our submission details.