Hugh Zhang
Graduate Fellow
PhD Student in EconCS
he/him
Contact Information
About
Hugh Zhang is a PhD student at Harvard, where he is advised by David Parkes and supported by the NSF Graduate Research Fellowships Program and the Kempner Graduate Fellowship. Previously, he was a software engineer at Asana during a gap year before college, studied Economics at Stanford, and he has worked at Google Deepmind and Meta AI. His main research interest is language modeling and search. Some of his past work includes CICERO, the first AI agent to achieve human-level performance in the game of Diplomacy. In Zhang’s spare time, he is and has been a lifelong Go player (seeing AlphaGo beat Lee Sedol was in fact the origin of his interest in AI). He also co-founded the Gradient, a digital magazine focusing on AI, and serves as its lead editor.
Research Focus
Zhang’s current research focuses on language models, search, and reinforcement learning. He is interested in language modeling and reinforcement learning. Recently, he has been interested in looking into methods for language models to perform search / planning by spending computational resources at inference time. These techniques (e.g. chain-of-thought reasoning or Monte-Carlo Tree Search) may have the potential to improve the capabilities for models tremendously beyond simply directly querying them for an answer.