Our Generative AI research is dedicated to advancing the core probabilistic foundations of machine learning, developing cutting-edge models and algorithms, and ultimately driving a revolution in artificial intelligence. Our work is structured around three core aspects: fundamental tools, theoretical understanding, and broad applications.

Fundamental Tools: Probabilistic tools lie at the heart of generative modeling, fundamentally shaping AI’s ability to learn, adapt, and create. Our research focuses on developing probabilistic frameworks that enhance AI’s capacity to understand complex environments, reason through intricate problems, and generate meaningful content.

Theoretical Understanding: We investigate the foundational principles of generative modeling, driven by a curiosity to uncover the mechanisms of generalization and disentangle the roles of data, model, and learning. By unraveling these intricacies, we aim to achieve utmost explainability and controllability in generative AI.

Applications: We apply our developed tools and theoretical insights to key domains such as computer vision and natural language processing, driving innovation and expanding the real-world impact of generative AI.