Reproduce “AuroraWorld: Structured 4D World Modeling for Exo-to-Ego Video Generation in Robot Manipulation” 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

Hao Xue, Peng-Fei Zhang, Ying Cheng, Lanyun Zhu, Lei Zhu, Heng Tao Shen

AuroraWorld synthesizes wrist-level videos from sparse external-camera footage using inferred scene geometry, motion trajectories, and selected keyframes to guide a video generator. On DROID, it reports PSNR of 17.87 versus 15.77 and FVD of 60.171 versus 70.073 for BWM. In ARX X5, policies average 44.7% success versus 58.7% with real views; cup stacking is an exception at 0/25.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.