Yizhou Zhang

I am a PhD student in Computing and Mathematical Sciences at Caltech, advised by Adam Wierman and Eric Mazumdar. My research tackles decision-making in multi-agent systems through game theory, reinforcement learning, and control, with a current emphasis on strategic risk aversion, robustness, and privacy in learning algorithms.

My recent work explores how to design provably convergent learning dynamics for general-sum Markov games, how risk-aware actor-critic methods shape strategic behavior, and how intrinsic regularization (e.g., KL terms) can deliver differential privacy in bandits and RLHF without explicit noise injection. I am broadly interested in decision-making at the intersection of economics, safety, and generative models for data-efficient policy learning.

Previously, I earned my B.Eng in Computer Science (Yao Class) from Tsinghua University. I spent time as a visiting undergrad student at Caltech through the Visiting Undergraduate Research Program (VURP) and as a research intern at the Shanghai Qi Zhi Institute working on platform economics.

When I am not coding, reading papers or proving convergence guarantees, I enjoy working out in the gym, playing basketball, cooking and listening to music.