RESEARCH PAPERYear 2026

Learning Latent Action World Models In The Wild

Quentin Garrido; Tushar Nagarajan; Basile Terver; Nicolas Ballas; Yann LeCun; Michael Rabbat

Classification

View four quadrants
Major category
WAMs
Architecture
Not applicable
Prediction paradigm
Not applicable
Source review status
Verified from primary sources

Category review. This work learns inverse-transition latent actions and latent forward dynamics, then trains a controller mapping candidate real actions into that latent space. CEM searches predicted trajectories toward a visual goal. This is an actionable world-model planning system; V-JEPA is the reusable encoder component. Reading evidence

AT A GLANCE

Contribution

A contribution summary has not been added yet.

Abstract

An abstract has not been added yet.

Affiliations

Not listed in the collection.

BibTeX