Reuse Your Rewards: Reward Model Transfer for Zero-Shot Cross-Lingual Alignment

arXiv:2404.12318v1 Announce Type: new
Abstract: Aligning language models (LMs) based on human-annotated preference data is a crucial step in obtaining practical and performant LM-based systems. However, multilingual human preference data are difficult to obtain at scale, making it challenging to extend this framework to diverse languages. In this work, we evaluate a simple approach for zero-shot cross-lingual alignment, where a reward model is trained on preference data in one source language and directly applied to other target languages. On summarization …

abstract alignment arxiv cross-lingual cs.cl data diverse framework however human language language models languages lms making multilingual practical reward model scale systems transfer type work zero-shot