Reproduce “Seeing Beyond the Next Step: World-Model-Guided Human-Like Navigation in Multi-Agent Scenes” 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
Seong-Eun Hong, JuYeong Hwang, Jinhyun Kim, Hanyoung Jang, HyeongYeop Kang
The authors train a navigation policy to choose motion anchors using preferences from two models: one predicts people’s motion, and the other predicts scene semantics. The models guide training but are removed at deployment, where the policy acts in one forward pass. Across five held-out Unity scenarios, it leads on success and social-region violations, but trails online planning on average.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.