A Zeroth-Order Paradigm for LLM Preference Alignment
Direct preference alignment methods are widely used to align large language models (LLMs) with human preferences because of their computational and memory efficiency. However, likelihood displacement motivates alternative ways to extract information from preference pairs with small likelihood margins. In this paper, we propose and analyze Comparison-based Preference Optimization (ComPO), a…
We haven't written up this one. arXiv cs.AI has the full story — the link below goes straight to it.