Livia Betti

prof_pic.jpg

Hi! My name is Livia!

I’m a PhD student at CU Boulder studying statistical and geospatial machine learning. I am advised by Esther Rolf. As of Fall 2025, my work is supported by the NSF-GRFP.

Before coming to CU, I completed by undergraduate degree in mathematics at the University of Rochester, where I did research on VC-dimension under the mentorship of Alex Iosevich. I’ve changed paths quite a bit since starting graduate school, and I am currently working in the intersection of data-centric AI and machine learning theory for geospatial applications. In Dyson’s classification, I am a frog.

Outside of my academic life, I enjoy running, backpacking, and board games :)

news

Jun 23, 2026 New preprint: What’s in an Earth Embedding?, an explainability analysis of location encoders, is out on arXiv!
Apr 17, 2026 New preprint: OT on the Map, on quantifying domain shifts in geographic space, is out on arXiv!
Jan 26, 2026 Attended AAAI 2026 and presented Mapping on a Budget!
Nov 03, 2025 Our paper Mapping on a Budget was accepted to AAAI 2026 Social Impact Track!
May 01, 2025 Attended ICLR 2025! Our paper Balancing quantity and representativeness in constrained geospatial dataset design was selected as an oral at the ML4RS workshop and presented as a poster at the Climate Change AI workshop.

selected publications

  1. mapping_on_a_budget.png
    Mapping on a Budget: Optimizing Spatial Data Collection for ML
    L. Betti, F. Sanni, G. Sogoyou, and 4 more authors
    In Proceedings of the AAAI Conference on Artificial Intelligence, 2026
    arXiv
    N/A Google Scholar citations
  2. ot_on_the_map.png
    OT on the Map: Quantifying Domain Shifts in Geographic Space
    H. Zhang, L. Betti, K. Klemmer, and 2 more authors
    arXiv preprint arXiv:2604.16220, 2026
    arXiv
  3. explainable_earth_embeddings.png
    What’s in an Earth Embedding? An Explainability Analysis of Location Encoders
    L. Betti, S. Ricke, I. Obadic, and 2 more authors
    arXiv preprint arXiv:2606.24997, 2026
    arXiv