1 paper
Zhongxi Chen, Shenqi Zong
Long-horizon robot manipulation requires policies to track completed subtasks and critical interaction events. However, existing memory mechanisms heavily rely on external models o…