Fetching the paper…

Weak-to-Strong Preference Optimization: Stealing Reward from Weak Aligned Model · Around