Zeyu (Steven) Zhang

I am a third-year Ph.D. student in the Department of Statistics and Data Science at Northwestern University. My research focuses on knowledge leakage in large language models, with particular emphasis on settings that involve complex reasoning and chain-of-thought processes. I study how and when sensitive or unintended information may be exposed during multi-step inference, and I am interested in principled methods for detecting, characterizing, and mitigating such leakage. In parallel, I work on fine-tuning language models using reinforcement learning–based approaches, including RLHF, Direct Preference Optimization (DPO), and related algorithms, with the goal of improving alignment, robustness, and controllability.

Outside of research, I am deeply committed to an active lifestyle. I follow a structured workout program and train almost every day, viewing physical discipline as a natural complement to intellectual rigor. I enjoy a wide range of sports, with basketball being my favorite, and I am particularly drawn to long-distance running for both its physical demands and its meditative rhythm. These activities keep me grounded, energized, and continuously motivated—both on and off the track.