Reproduce “AgentDreamer: Agentic 3D Motion Reasoning for Text-Controllable Physics Simulation” 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
Mijeong Kim, Suho Park, Junha Park, Bohyung Han
AgentDreamer uses an agent to turn text instructions into 3D motion plans, which guide a Material Point Method simulator as it fits parameters. Its plans use parametric functions to produce trajectories, rather than relying on motion extracted from video. Across eight scenes and 20 instructions, it ranked first in 73.6% of rankings; these measure text–motion alignment, not physical correctness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.