Tatsu Hashimoto

• Program

  AI Safety Science

• Institution

  Stanford University

• Location

  USA

Dr. Tatsu Hashimoto is an Assistant Professor in the Computer Science Department at Stanford University. His research leverages statistical methods to enhance the robustness and trustworthiness of machine learning systems, particularly in complex environments involving large language models. Dr. Hashimoto is focused on addressing fundamental challenges in machine learning and natural language processing, including long-tail behavior, system robustness under changing real-world conditions, robust language understanding, and fairness to prevent harmful predictions from unreliable correlations.  
Before joining the faculty at Stanford, Dr. Hashimoto was a postdoctoral researcher at Stanford, working with John C. Duchi and Percy Liang on analyzing tradeoffs between average and worst-case model performance. He earned his PhD from MIT, advised by Tommi Jaakkola and David Gifford, and completed his undergraduate degree in statistics and mathematics at Harvard University under the guidance of Edoardo Airoldi.

Dr. Hashimoto’s project aims to develop new safety metrics capable of identifying emergent, unsafe AI capabilities proactively.