DeepJEPA: Scaling World Models from Within

What happened

arXiv:2610.00368v1 Announce Type: new Abstract: World-model planners typically scale outward by rolling farther, sampling more trajectories, or optimizing longer, while assigning the same computation to every imagined transition. Its additional computation concentrates at contact onset and sustained object interaction, where latent corrections can change which candidates enter the planner's elite set and which action is selected.

Across five visual-control settings, DeepJEPA improves or matches the strongest fixed-depth planner while averaging only 1.00-1.26 updates per transition. Representation probes further show that improved planning does not require uniformly better object-state decodability. DeepJEPA therefore reframes world-model scaling as a problem of allocating internal computation where it can change the planner's decision: think deeper at decision-critical transitions instead of making every rollout uniformly deeper or longer.