Key points are not available for this paper at this time.
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.
Building similarity graph...
Analyzing shared references across papers
Loading...
Dang et al. (Tue,) studied this question.
www.synapsesocial.com/papers/68e61b7fb6db6435875ae584 — DOI: https://doi.org/10.48550/arxiv.2407.02552
John Dang
Arash Ahmadian
Kelly Marchisio
Building similarity graph...
Analyzing shared references across papers
Loading...