Reproduce “Flow-ERD: Agent-type Aware Flow Matching With Entropy-Regularized Distillation for Diverse Traffic Simulation” 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
Seulbin Hwang, Kiyoung Om, Dae Jung Kim, Jinhan Lee
Flow-ERD generates continuous traffic actions, executes them with type-specific kinematics, and fine-tunes on closed-loop rollouts using entropy-regularized distribution matching to retain less common behaviors. On WOSAC 2025 test, it ranked first with RMM 0.7878; on validation, RMM rose from 0.7836 to 0.7869 while CPD fell from 0.1858 to 0.1828, indicating most pretrained diversity remained, not universal validity.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.