Scholar
Ziming Liu
Google Scholar ID: QeXHxlIAAAAJ
MIT
Citations & Impact
Citations
5,670
H-index
21
i10-index
27
Publications
20
Co-authors
44
list available
Publications
Resume (English only)
Academic Achievements
- - Publications:
- - KAN: Kolmogorov-Arnold Networks
- - Seeing is believing: Brain-inspired modular training for mechanistic interpretability
- - Scientific discovery in the age of artificial intelligence
- - The Clock and the Pizza: Two Stories in Mechanistic Explanation of Neural Networks
- - Omnigrok: Grokking Beyond Algorithmic Data
- - Poisson Flow Generative Models
- - Machine Learning Hidden Symmetries
- - Machine Learning Conservation Laws from Trajectories
Research Experience
- - Postdoc at Stanford & Enigma, collaborating with Prof. Andreas Tolias
- - Conducted research at MIT
Education
- - Ph.D.: Massachusetts Institute of Technology, advised by Prof. Max Tegmark
- - B.S.: Peking University
Background
- - Research Interests: Intersection of AI and Science
- - Professional Field: Science of AI, Science for AI, AI for Science
- - Background: Currently a postdoc at Stanford & Enigma, working with Prof. Andreas Tolias. His research philosophy is inspired by John Hopfield.
Miscellany
- - Media Coverage: Quanta magazine and other media outlets have covered all three branches of his research.
- - Latest Blog Posts:
- - Philosophical thoughts on Kolmogorov-Arnold Networks
- - Symbolic Regreesion? Structure Regression!
- - A Good ML Theory is Like Physics -- A Physicist's Analysis of Grokking
