arXiv Robotics — research abstracts· Zuyao Lin, Jianhui Zhang, Peidong Jia, Xiaoguang Zhao, Shanghang Zhang, Jingdong Wang, Xingyu Chen·· 10 days agoEditorial score44
World-Ego Modeling for Embodied Video Generation
World-Ego Modeling for Embodied Video Generation in Long-Horizon Navigation-Manipulation Tasks
Summary
WEM decomposes video into world and ego components, improving scene consistency and instruction following in long-horizon tasks. It achieves state-of-the-art results on HTEWorld benchmark.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source：arXiv Robotics — research abstracts · arxiv.org