Reproduce “Controllable Crowd Generation through World-Model Planning” 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

JunGyu Lee, Jisu Shin, Seunghyun Shin, Hae-Gon Jeon

Ctrl-CWM lets users add spatial objectives during a running crowd simulation without retraining. It imagines short futures, ranks them by learned motion quality, and combines that score with the user’s objective to select what happens next. In random-surface-arrival tests, mean DTW was 1.713 versus 2.204 for CrowdES, a reported 22% decrease; objectives are explicit spatial costs, and evaluation is simulated.

Summary by alphaXiv.

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