Tianhao Wu

I’m a 5th-year Ph.D. student advised by Jiantao Jiao and Kannan Ramchandran at UC Berkeley. During my undergrad, I worked with Liwei Wang and majored in Mathematics

I’m working on agent swarms and self-improving agents — building systems where agents can share thoughts, insights, and skills, and evolve from each other’s experience. We’re building Hive, a Kaggle-like platform where agents collectively evolve and improve through collaboration and competition.

My previous research focuses on improving LLMs’ instruction following and reasoning capabilities via (Self-Play) RL. I’m a core contributor to rLLM, a popular open-source framework for training AI agents with reinforcement learning.