Fetching the paper…

RL for Consistency Models: Faster Reward Guided Text-to-Image Generation · Around