Flow-ERD is a two-stage multi-agent traffic simulator designed to optimize realism and diversity jointly. Its Agent-Type Aware Flow Matching (AFM) combines the multimodal expressiveness of flow matching with type-specific kinematic execution, aiming to preserve fine-grained behavioral diversity while maintaining motion consistency for different agent categories. Entropy-Regularized Distillation (ERD) then fine-tunes the closed-loop rollout distribution using an entropy-regularized reverse-KL objective. The paper reports that Flow-ERD ranks first on the WOSAC test benchmark and dominates the realism-diversity Pareto front among reproducible baselines, using a log-free diversity metric alongside standard realism scores.
No heat snapshots are available in the last 24 hours.