Reproduce “Hierarchical World Models with Implicit Dynamics” 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
Gaoyue Zhou, Yvonne Wu, Zichen Jeff Cui, Nicolas Ballas, Mido Assran, Lerrel Pinto, Yann LeCun
ID-WM separates choosing intermediate states from choosing actions to make long-horizon planning in world models manageable. A diffusion model proposes nearby subgoals, and an action-conditioned model searches for actions to reach them, with planning repeated locally. In state-based PushT random-goal trials, it succeeds in 77% versus 21% for single-level world-model MPC; this result is limited to the tested benchmark setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.