By predicting future states, latent world models provide rich self-supervision for end-to-end driving, but often inherit dense features from pretrained visual encoders as their world-state interface. This creates a mismatch: visual pretraining produces information-rich, spatially dense features, but their adaptation fo...
Yi Gu, Hao-Yu Wang, Shao-Cheng Yan et al.· 0 citations
RISE is proposed, a system-level adaptive imagination framework that makes sequential sequential Roll/Stop decisions according to the expected planning benefit of continued rollout, with additional transfer results supporting its plug-in generality across WAM architectures.
Hong-Bo Lu, Liang Yao, Chenghao He et al.· 4 citations
DroneWAM adopts a JEPA-based architecture to model future states directly in representation space, avoiding the cost of explicit future image generation and reduces the average prediction depth from 8 to 4.58 while improving trajectory accuracy, demonstrating that predictive computation can be allocated more effectivel...
Liang Yao, Fan Liu, Hong-Bo Lu et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.