Pre-Trained Policy Discriminators are General Reward Models Paper β’ 2507.05197 β’ Published 1 day ago β’ 28