Reproduce “Embodied Graph: A Relational World Model for Long-Horizon Tasks” 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

Zongyuan Wu, Xin Wang, Tongtong Feng, Yuan Meng, Zeyang Zhang, Ling Feng, Wenwu Zhu

RWM uses an online-updated graph of task-relevant entities and predefined relations as the state for world-model imagination, predicting how node states and interactions change. Its graph model guides action selection in embodied and web tasks. In Minecraft, it achieved 6.94 ± 2.12% diamond success after 5 million steps, versus 0% for four baselines, under the paper’s specified setup.

Summary by alphaXiv.

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