1 paper · 1 filter
Hengyan Liu, Wenlve Zhou, Bo Yue +7
Reactive vision--language--action (VLA) models struggle with long-horizon manipulation when visually similar observations can correspond to different actions depending on the task…