Glossary · AI & ML
World model
In brief
A world model is a learned simulator of the environment used to plan or predict consequences of actions. Instead of (or in addition to) a real-world rollout, the agent imagines forward in its world model, picks the best plan, then acts.
World models are how AI agents do model-based planning. The classic loop: encode the current observation into a latent state, predict how that state evolves under candidate actions, score the rollouts, and pick the best action. The big advantage over model-free RL is sample efficiency — you can plan in your head a lot more cheaply than in the real world.
For humanoids, world models are most relevant for high-level planning over longer horizons (cleaning a room, navigating a building). NVIDIA Cosmos is one public world-model platform aimed at robotics; Wayve and several academic groups have published competitive systems.
Related terms
See it in the wild
Browse robots and brands using these techniques
Glossary entries are upstream. The catalog is where the implementations live.