Reproduce “DeXplicit: Making Physical State Explicit in Dexterous Manipulation World Models” 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
Adam Lee, Mutian Xu, Yukang Cao, Shulin Tian, Ziwei Liu
DeXplicit predicts how a hand’s planned motion changes an object’s metric 3D state, then uses that forecast to guide video synthesis. It models hand joints and object motion, transports visible appearance, and leaves uncertain regions to the renderer. On HOT3D, moved-point error is 54.34 mm versus PointWorld’s 90.47 mm; it assumes initial metric 3D geometry and supplied hand trajectories.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.