Fetching the paper…

Diffusion Model as a Noise-Aware Latent Reward Model for Step-Level Preference Optimization · Around