Reproduce “Lifting Embodied World Models for Planning and Control” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Alex N. Wang, Trevor Darrell, Pavel Izmailov, Yutong Bai, Amir Bar
The authors introduce waypoint-based planning for human motion with a frozen world model. A learned policy maps four image-plane targets for the pelvis, head, and hands to joint actions. On 128 Nymeria validation tasks, Lifted CEM reaches 0.374 m all-joint MJE versus 0.616 m for direct PEVA CEM in model-generated rollouts, where a goal joint is initially visible.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.