Science, World models and Intelligence

Generative world models for science and health.

Explore our research
SWI / RESEARCH PERSPECTIVE
Conceptual field visualization: a family of waves represents a changing physical system.a learned representation of change
Observationspartial · noisy · multimodal
infer generate
World modelstructure · dynamics · uncertainty
One model. Two directions.Conceptual illustration

OUR SHARED QUESTION

What can we learn about the world
from what we can observe?

RESEARCH

Across pixels, physics,
and physiology.

A shared foundation in generative modeling.
Two connected worlds of application.

THE IDEA THAT CONNECTS US

Simulate forward.
Reason backward.

A world model is more than a generator. It is a prior for reconstruction, a simulator for physics, and a model of physiological dynamics.

We bring these perspectives together to infer what produced an observation—and predict what might happen next.

Our approach

WORKING TOGETHER

Depth in methods.
Honesty in applications.

We care about rigorous ideas and systems that work beyond the benchmark. Our research connects machine learning with questions that matter to scientists and clinicians.