Embodied Intelligence
Agents that perceive, reason, and act in open worlds.
Intelligence, grounded in reality.
Four research territories, connected by a shared ambition: to build intelligence that can perceive, predict, and participate in the physical world.
World Models
Predictive representations for dynamic, interactive environments.
Spatial & Physical Intelligence
Geometry and physical priors for grounded intelligence.
Multimodal Intelligence
Intelligence across vision, language, sound, and space.
Perception. Action.
Our work connects perception, simulation, reasoning, and control—because physical intelligence is not a single model, but a closed loop with the world.
Observe
Model
Reason
Act
Groundedintelligence