Reproduce “Fast-JEPA: Advancing the Success-Latency Frontier for Latent World Models” 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

Yao Zhang, Bin Yu, Shijie Lian, Haishan Liu, Changti Wu, Yu Wu, Kai Chen

Fast-JEPA is a visual controller that reduces planning latency by using intermediate predictions to construct goal-directed plans and direct terminal predictions to compare completed candidates. Its latent world model scores predicted endpoints against a goal, without assessing path constraints. Across four simulated tasks, it reports 96.50% average success at 10.88 ms, versus INTACT’s 95.58% at 26.19 ms under the stated benchmark protocols.

Summary by alphaXiv.

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