Announcement_4

New preprint: REFINE-DP presents a reinforcement learning fine-tuning framework for diffusion-policy-based humanoid loco-manipulation, achieving 90%+ simulation success and strong real-world performance on door traversal and object transport.