1 paper
Xingyu Ding, Yuzhong Zhao, Chunhai Zhao +3
Long-horizon robotic manipulation requires vision-language-action (VLA) models to track scene states and their evolution beyond the current observation. However, simply conditionin…