Sarah Wiegreffe

Sarah Wiegreffe is an Assistant Professor in the Department of Computer Science at the University of Maryland, affiliated with the CLIP laboratory. She holds a PhD from Georgia Institute of Technology (2022), advised by Prof. Mark Riedl. Previously, she was a postdoctoral researcher at the Allen Institute for AI (Ai2) and a member of H2Lab at the University of Washington, advised by Ashish Sabharwal and Hannaneh Hajishirzi.

Her research focuses on mechanistic interpretability and transparency of language models, aiming to improve reliability, safety, and performance through natural language explanations. Notable contributions include the Mechanistic Interpretability Benchmark (MIB) and work on explaining model noncompliance.

Awards include the Ai2 Outstanding Intern Award, Rising Star in Machine Learning/EECS/Generative AI, and recognition as an Outstanding Area Chair (ACL 2023, NeurIPS 2023). She has advised students like Alec Bunn and contributed to workshops like the Actionable Interpretability at ICML 2025.

Education: PhD in Computer Science, Georgia Tech (2022); Master’s/Undergraduate degrees not explicitly stated. Active in organizing conferences (e.g., NAACL 2021 Publicity Co-Chair) and teaching (e.g., graduate seminar CMSC848R on LM interpretability).