Multi-Foundation Intelligence Lab (MF-Lab)
University of Washington
The Multi-Foundation Intelligence Lab (MF-Lab) is a research group at the University of Washington dedicated to understanding how intelligent systems can build structured internal representations of the world and use them to make robust, trustworthy decisions. We develop model-based and predictive systems that can anticipate future outcomes, reason under uncertainty, and adapt to changing conditions.
Our work integrates multimodal perception, generative modelling and decision‑making, and emphasises reliability, controllability and rigorous evaluation. By bridging advances in modern machine learning with the constraints of real‑world deployment and human‑centred use, we aim to bring foundational research into practical, trustworthy AI applications.
Model‑based and predictive intelligence grounded in multimodal and real‑world environments
Our current research topics are as follows:
World Modelling
- Learning world representations and latent dynamics for prediction and planning
- Generative and predictive modelling of complex environments
- Safe and controllable diffusion‑based world models
Multimodal & Embodied Intelligence
- Vision–language–action integration for situated intelligence
- Human‑aware perception and multimodal reasoning
- Embodied agents interacting in physical and virtual environments
Safety, Reliability & Systems
- Safety‑aware training objectives and evaluation methods
- Failure modes and uncertainty modelling under distribution shift
- Decision‑making and deployment in safety‑critical systems
If you are interested in our research or potential collaborations, please reach us at zhiqics@uw.edu.
Join MF‑Lab
We welcome highly motivated Ph.D. students and collaborators who share our passion for building trustworthy model‑based intelligence.
See OPEN POSITIONS for current opportunities.
Contact
Zhi‑Qi Cheng
Assistant Professor
University of Washington
Email: zhiqics@uw.edu