LeWAM End-to-End JEPA for Faster World and Action Modeling
October 12, 2026
LeWAM uses Joint-embedding predictive architectures to learn latent world models without pixel reconstruction, employing action flow matching and an isotropic Gaussian constraint. In testing, it increased success rates from 28.6% to 89.7% on contact-rich tasks and achieved 32.3x faster planning speeds.
HOW THIS AFFECTS YOU
●
builderYou can deploy significantly faster planning models for robotics and contact-rich control tasks.
●
researcherYou can utilize joint training objectives over latent state prediction and action flow matching for improved control.