Haidong Cao

HiMoE-VLA: Hierarchical Mixture-of-Experts for Generalist Vision–Language–Action Policies

Unleashing 2D Rewards for Human Preference Aligned Text-to-3D Generation via Preference Score Distillation