Reproduce “WorldMirror: Agentic Real-to-Sim Generation of Interactive Worlds from In-the-Wild Robot Demonstrations” 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
Tianling Xu, Weijie Wan, Honghao Su, Kailun Su, Junyuan Tang, Huichi Zhou, Haozheng Zhang, Qi Xiong and 4 more
WorldMirror turns real-world manipulation videos and recorded robot motion into editable simulation scenes for replay. It coordinates scene context, task objects, and the robot in a shared camera-anchored frame, then verifies updates and uses failure feedback to guide revisions. In Isaac Sim, evaluated DROID demonstrations achieved 100% replay success versus 40% for GPT-6 Astra, showing simulation replay, not real-world transfer.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.