Game-Theoretic Diffusion Models

Abstract

AI systems have significant potential to drive progress in both industrial and scientific domains, particularly in areas such as AI-assisted drug discovery. However, conventional AI models often operate within static and single-agent decision-making frameworks, limiting their adaptability to complex, real-world challenges.

In this talk, I will discuss the fundamental limitations of these existing frameworks and present algorithmic approaches designed to address them. I will then introduce a research project focused on integrating multi-agent paradigms and dynamic modeling, aiming to enhance the practical effectiveness of future AI systems by bridging the gap between theoretical foundations and real-world complexities.

Please click here for participating the talk online.

Download