Value-Incentivized Preference Optimization: A Unified Approach to Online and Offline RLHF

Open in new window