Alessandro Favero

EPFL - École Polytechnique Fédérale de Lausanne

profile_pic.png

I’m a final-year PhD student at EPFL in Switzerland, working with Matthieu Wyart and Pascal Frossard. This fall, I will join the University of Cambridge as a Research Associate and Physics-AI Fellow.

My research aims to deepen the scientific understanding of AI and deep learning, often drawing on methods from statistical physics. I am currently focused on generative models – in particular, continuous and discrete diffusion models – and the science of post-training, including large-scale model editing and merging.

Previously, I interned as an Applied Scientist in the Fundamental Research team at Amazon’s AWS AI Labs in the Bay Area, and completed my studies in theoretical physics at Sorbonne, Politecnico di Torino, SISSA, and ICTP.

news

Apr 08, 2025 Talk on compositionality in diffusion models at the Theory+AI event, Perimeter Institute.

selected publications

  1. compositional_diffusion.png
    How compositional generalization and creativity improve as diffusion models are trained
    Alessandro Favero*, Antonio Sclocchi*, Francesco Cagnetta, Pascal Frossard, and 1 more author
    International Conference on Machine Learning (ICML), PMLR, 2025
  2. phase.png
    A phase transition in diffusion models reveals the hierarchical nature of data
    Antonio Sclocchi, Alessandro Favero, and Matthieu Wyart
    Proceedings of the National Academy of Sciences (PNAS), 2025
  3. multimodal.png
    Multi-modal hallucination control by visual information grounding
    Alessandro Favero, Luca Zancato, Matthew Trager, Siddharth Choudhary, and 4 more authors
    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024
  4. task_arithmetic_2.png
    Task arithmetic in the tangent space: Improved editing of pre-trained models
    Guillermo Ortiz-Jimenez*Alessandro Favero*, and Pascal Frossard
    Advances in Neural Information Processing Systems (NeurIPS), Oral presentation (top 0.54%) , 2023