Domain
World Models
Learned simulators, action-conditioned video prediction, and the JEPA counterargument.
5 of 5 modules published
Six paradigms share one name: latent dynamics, decoder-free latent, generative video, JEPA, unified world-action, and symbolic.
Dreamer, TD-MPC2, and DayDreamer: compact learned dynamics for imagination-based control.
Cosmos, Genie, and GR-2: action-conditioned video prediction and the conditioning-strength problem.
V-JEPA 2 and LeCun's case that prediction in representation space beats pixel generation.
RoboGen, Holodeck, and RoboCasa: generated content inside real physics engines beats generated dynamics.