机器人 / 具身智能

CLAM: Continuous Latent Action Models for Robot Learning from Unlabeled Demonstrations

CLAM：基于未标记演示的连续潜在动作模型用于机器人学习

Anthony Liang, Pavel Czempin, Matthew M. Hong, Yutai Zhou, Jingzhen Wang, Erdem Biyik, Stephen Tu

机构 * Department of Computer Science, University of Southern California（南加州大学计算机科学系） ； Department of Electrical and Computer Engineering, University of Southern California（南加州大学电气与计算机工程系） 纠错

专题命中 机器人学习 ：robot learning(title)；manipulation(abstract)；分类 cs.RO、cs.AI、cs.LG

AI总结 CLAM通过连续潜在动作标签和联合训练动作解码器，有效解决复杂连续控制任务中未标记数据的学习问题，实现在DMControl和MetaWorld等基准及真实机器人上的性能提升。

Comments Latent Action Models, Self-supervised Pretraining, Learning from Videos

Journal ref IEEE/RSJ International Conference on Intelligent Robots and Systems 2026