arXiv Robotics — research abstracts· Qi Chen, Yuwei Wu, Quan Chen, Kunpeng Gao, Jin Chen, Che Sun, Zhi Gao, Yunde Jia·· 11 hours agoEditorial score42
One-shot Human Video Guidance for VLA Adaptation
One-shot Human Video-Guided Adaptation for Vision-Language-Action Models
Summary
CapPro enables VLA models to adapt to new tasks with a single human video. It uses phase-aware representations and video-conditioned adaptation to improve performance on unseen tasks without requiring robot demonstrations.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source：arXiv Robotics — research abstracts · arxiv.org