1 paper
Kangjun Guo, Haichao Liu, Yanji Sun +3
Developing a unified policy for multi-task robotic manipulation remains challenging due to policy degradation from task interference and negative transfer. In this work, we propose…