Avoiding mode collapse in diffusion models fine-tuned with reinforcement learning

Re-assign community