Reproduce “ActWorld-JEPA: Action-Conditioned JEPA World Models Meet Latent Planning 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

Ziying Song, Shengkai Zhang, Caiyan Jia, Lei Yang, Lin Liu, Yuchen Liu, Ziyang Liu, Zirui Li and 5 more

ActWorld-JEPA reranks proposed trajectories by predicting a compact future representation for each candidate. PC-JEPA conditions a latent rollout on each plan, while R³ uses the relation between plan intention and predicted future to produce a bounded score correction. On NAVSIM v2 navtest, full R³ scores 91.6 EPDMS versus 90.1 for Score Replacement, supporting this design in a non-reactive benchmark.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.