RLHF Can Speak Many Languages: Unlocking Multilingual Preference Optimization for LLMs
Dang, John, Ahmadian, Arash, Marchisio, Kelly, Kreutzer, Julia, Üstün, Ahmet, Hooker, Sara
–arXiv.org Artificial Intelligence
Preference optimization techniques have become a standard final stage for training state-of-art large language models (LLMs). However, despite widespread adoption, the vast majority of work to-date has focused on first-class citizen languages like English and Chinese. This captures a small fraction of the languages in the world, but also makes it unclear which aspects of current state-of-the-art research transfer to a multilingual setting. In this work, we perform an exhaustive study to achieve a new state-of-the-art in aligning multilingual LLMs. We introduce a novel, scalable method for generating high-quality multilingual feedback data to balance data coverage. We establish the benefits of cross-lingual transfer and increased dataset size in preference training. Our preference-trained model achieves a 54.4% win-rate against Aya 23 8B, the current state-of-the-art multilingual LLM in its parameter class, and a 69.5% win-rate or higher against widely used models like Gemma-1.1-7B-it, Llama-3-8B-Instruct, Mistral-7B-Instruct-v0.3. As a result of our study, we expand the frontier of alignment techniques to 23 languages covering half of the world's population.
arXiv.org Artificial Intelligence
Jul-2-2024
- Country:
- South America
- North America
- Dominican Republic (0.04)
- United States
- Washington > King County
- Seattle (0.04)
- California > Santa Clara County
- Palo Alto (0.04)
- Washington > King County
- Canada > Ontario
- Toronto (0.04)
- Europe
- Spain (0.04)
- Monaco (0.04)
- Italy > Tuscany
- Florence (0.04)
- Ireland > Leinster
- County Dublin > Dublin (0.04)
- France > Provence-Alpes-Côte d'Azur
- Bouches-du-Rhône > Marseille (0.04)
- Faroe Islands > Streymoy
- Tórshavn (0.04)
- Estonia > Tartu County
- Tartu (0.04)
- Asia
- Singapore (0.04)
- India (0.04)
- Myanmar > Tanintharyi Region
- Dawei (0.04)
- Middle East
- Jordan (0.04)
- Israel (0.04)
- UAE > Abu Dhabi Emirate
- Abu Dhabi (0.04)
- Japan > Honshū
- Chūbu > Toyama Prefecture > Toyama (0.04)
- Africa > Zambia
- Southern Province > Choma (0.04)
- Genre:
- Research Report (0.64)
- Technology: