Juan Pablo Vigneaux
AI and Mathematics

I am currently in the job market!
From October 2021 to December 2024, I served as an Olga Taussky and John Todd Instructor in Mathematics at Caltech, mentored by Matilde Marcolli. We co-organized the Information, Geometry, and Physics Seminar.
Previously, I was a postdoctoral researcher at the Max Planck Institute for Mathematics in the Sciences. In 2019, I obtained my Ph.D. in Mathematics at the University of Paris under the supervision of Daniel Bennequin.
I’m currently working on interpretability of transformer-based LLMs, with a particular focus on the geometry of latent spaces and on compositionality. For further details, see the research section.
More broadly, I’m interested in mathematical aspects of information theory, particularly in connection with category theory and geometry (metric geometry, geometric measure theory, …). My published work can be divided in three different areas:
- Characterization of information measures, mainly information topology.
- Information dimension and measures with geometric structure.
- Magnitude and diversity.
Here’s my academic CV.
latest posts
Aug 2, 2024 | Kolmogorov's axiomatization of probability |
---|---|
May 4, 2024 | An invitation to ungrade |