Andre He

I’m a third-year Ph.D. student in the Language Technologies Institute at CMU, co-advised by Sean Welleck and Daniel Fried. Before coming to CMU, I completed my undergraduate studies at UC Berkeley. I had the privilege of working with Nicholas Tomlin and Dan Klein in the Berkeley NLP Group and with Kuan Fang and Sergey Levine in RAIL. I also completed an internship at AWS, where I was mentored by Nathaniel Weir and Kaj Bostrom.

I study how language models learn to solve problems and how they organize the computation needed to solve them. I’m interested in both how models discover useful strategies through exploration and feedback and how they carry out the reasoning process itself. This means studying what intermediate representations they use, which decisions they make, and where additional computation helps. My recent work explores these questions through reinforcement learning, diffusion language models, and automatically generated reasoning environments.

See my publications and CV for more details.