Jiayi Zhang

I am currently a PhD student at University of Chicago’s Data Science Insitute, advised by Prof. Chenhao Tan and affiliated with the broader Communication and Intelligence (C&I) group.

Previously, I spent two years as a PhD student at Northeastern University, building human-like LLM-based user simulators to bridge RL training with real-world interaction. Before that, I completed my B.S. with Honors in Computer Science at the University of Michigan, with minors in Mathematics and Linguistics. There, I had the privilege of being advised by Prof. Lu Wang, and of working closely with Prof. Zhijing Jin and Prof. Rada Mihalcea on NLP for social good.

My research interests lie in building LLM-based systems that achieve bidirectional alignment with humans and can co-improve over the long term in challenging domains. Currently, I’m particularly interested in:

• Improving how LLMs model human behavioral and conversational patterns, with the goal of building more faithful user simulators that can serve as training environments and probe grounded social interactions (aligning AI $\rightarrow$ Humans);
• Understanding and explaining complex model behaviors in human-interpretable ways (aligning Humans $\rightarrow$ AI);
• LLMs for Scientific Discovery.