Reproduce “WAM-Sim0.1: World Anchored Memory Meets Online RL for Consistent and Realistic Driving 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
Jiahao Cui, Jiaye Li, Wenhao Zhang, Yaoyi Li, Shan Luan, Quanhui Tang, Jianhua Han, Lin Cheng and 2 more
WAM-Sim generates long-horizon driving videos using persistent memories for road geometry, location-linked scene appearance, and tracked participants, alongside online reinforcement learning. In self-collected 60-second rollouts, it reports higher subject coherence and scene similarity and lower FVD than a comparison retaining map conditioning but removing scene and instance memories. This result supports the combined memories under the tested setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.