Michael Pearce
Bio
Michael Pearce is a researcher at Goodfire, an AI interpretability research company focused on mechanistic understanding of neural networks. He holds a PhD in Theoretical Physics from Stanford University, where his dissertation examined the dynamics of evolving populations and complex ecosystems under advisor Daniel Fisher, and a B.S. in Physics from MIT where he worked with Alan Guth on early universe cosmology. He participated in MATS 6.0 (Summer 2024), mentored by Lee Sharkey, where he developed methods to decompile interpretable transformer architectures, used Meta-SAEs to identify interpretable features within SAE latents, and applied information-theoretic approaches including minimal description length to guide SAE hyperparameter selection. At Goodfire, his applied research has included co-authoring work on using interpretability to identify novel Alzheimer's biomarkers and exploring phylogenetic structures in the Evo 2 genomic foundation model. He is based in San Francisco.
Links
- Personal Website
- -
- Twitter / X
- -
- LessWrong
- -
Grants
No grants recorded.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Mar 22, 2026, 11:37 PM UTC
- Created
- Mar 20, 2026, 3:00 AM UTC