Reproduce “Efficient World-Model Planning via Physics-Guided Search Distillation” 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

Hongyu Zhang, Xiaoshuai Hao, Heye Huang, Qingwen Meng, Wenjun Shi, Jianao Xu, Xin Hu, Xuanru Jin and 2 more

PGSD improves planning by checking how candidate paths execute, then repairing cases where none is road-compliant. It selects feasible paths using a controller, and otherwise predicts and verifies one lateral correction from future-aware features and execution feedback. On NAVSIM v2 navhard, it scores 38.96 EPDMS versus WorldDrive’s 27.72; this benchmark uses map-based execution checks, not vision alone or real-world driving.

Summary by alphaXiv.

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