1 paper
Yijing Zhou, Qiwei Liang, Sitong Zhuang +5
We present AEM, an Action-Effect Memory pretraining framework for robot manipulation that learns compact temporal representations from vision-action history. Unlike prior robot rep…