✳flâneur — a map of the web's best reading

LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels

le-wm.github.io · 1,074 words · saved by 3 readers

End-to-end joint-embedding predictive architecture from pixels.

LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels LeWorldModel : Stable End-to-End JEPA from Pixels Lucas Maes* 1 , Quentin Le Lidec* 2 , Damien Scieur 1,3 , Yann LeCun 2 , Randall Balestriero 4 , 1 Mila & Université de Montréal, 2 New York University 3 Samsung SAIL 4 Brown University * Equal Contribution Paper Code Data & Checkpoints Abstract Joint Embedding Predictive Architectures (JEPAs) offer a compelling framework for learning world models in compact latent spaces, yet existing methods remain fragile, relying on complex multi-term losses, exponential mov

Explore this link on the map →

saved by

related reading