机器人 / 具身智能

The Temporal Trap: Entanglement in Pre-Trained Visual Representations for Visuomotor Policy Learning

Nikolaos Tsagkas, Andreas Sochopoulos, Duolikun Danier, Chris Xiaoxuan Lu, Oisin Mac Aodha

专题命中 机器人学习 ：分类 cs.RO、cs.AI、cs.CV；robot learning(comments)

Comments This submission replaces our earlier work "When Pre-trained Visual Representations Fall Short: Limitations in Visuo-Motor Robot Learning." The original paper was split into two studies; this version focuses on temporal entanglement in pre-trained visual representations. The companion paper is "Attentive Feature Aggregation."