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

OSLM
Re-assign community