Seeing Beyond the Next Step: World-Model-Guided Human-Like Navigation in Multi-Agent Scenes
Overview
This research presents a method for human-like navigation in multi-agent scenes by using world models at training time to distill foresight into a fast policy, removing them at deployment. The approach pairs a discrete visual action prior with two task-aligned world models—one for nearby social dynamics and one for future scene semantics—to score candidate actions by their predicted consequences. Through group-relative preference distillation, the policy is refined so that at deployment it acts in a single forward pass without online rollouts. Across five held-out evaluation scenarios, the method achieves navigation that is safer, more socially and group-coherent, and more goal-directed than strong reactive and learned baselines, with naive observers rating the resulting motion as more human-like.
Generated from attributed reports · Updated 2 days ago
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
No current evidence-backed claims. Missing information remains not reported.
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsSignalSeeing Beyond the Next Step: World-Model-Guided Human-Like Navigation in Multi-Agent Scenes
This research presents a method for human-like navigation in multi-agent scenes by using world models at training time to distill foresight into a fast policy, removing them at deployment. The approach pairs a discrete visual action prior with two task-aligned world models—one for nearby social dynamics and one for future scene semantics—to score candidate actions by their predicted consequences. Through group-relative preference distillation, the policy is refined so that at deployment it acts in a single forward pass without online rollouts. Across five held-out evaluation scenarios, the method achieves navigation that is safer, more socially and group-coherent, and more goal-directed than strong reactive and learned baselines, with naive observers rating the resulting motion as more human-like.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).