1 paper
Gongxin Yao, Yongsheng Zhao, Jiayin Deng +4
Vision-Language-Action (VLA) models based on generative frameworks, such as Flow Matching, have recently achieved impressive performance in robotic manipulation. Unlike determinist…