1 paper
Yudong Liu, Yuan Li, Zijia Tang +12
Dual-system Vision-Language-Action (VLA) models achieve state-of-the-art robotic manipulation but are bottlenecked by the VLM backbone, which must execute at every control step whi…