Reproduce “EvoAD: Self-Evolving Autonomous Driving via Latent World-Model Guided Post-hoc Reflection” 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
Junjie Yang, Qiming Zhang, Yuyan Li, Shuqi Shen, Yusong Huang, Yanggang Sheng, Wenpeng Xu, Ziyi Shi and 7 more
EvoAD lets a deployed driving planner refine later decisions using post-execution feedback and scene-matched memory. A latent world model predicts compact future features that help evaluate candidate trajectories, while a refiner makes corrections without retraining the base planner. On NAVSIM v1, PDMS rises from 88.3 to 93.4 at round five; results are benchmark evidence, not on-road validation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.