1 paper
Fan Du, Feng Yan, Jianxiong Wu +8
Flow-based vision-language-action (VLA) policies offer strong expressivity for action generation, but suffer from a fundamental inefficiency: multi-step inference is required to re…