Reproduce “Latent-WAM: Latent World Action Modeling for End-to-End Autonomous Driving” 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
Linbo Wang, Yupeng Zheng, Qiang Chen, Shiwei Li, Yichen Zhang, Zebin Xing, Qichao Zhang, Xiang Li and 8 more
Latent-WAM trains a camera-based driving planner to encode road geometry and scene changes, using its future predictor only during training. Geometric distillation and multi-step prediction shape the encoder; inference uses the encoder and trajectory decoder. On NAVSIM v2, it scores 89.3 EPDMS, 3.2 points above the best perception-free method in the comparison, a benchmark result rather than evidence of safety.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.