Alessandro Favero
EPFL Institute of Physics
BSP 512 (Cubotron UNIL)
Rte de la Sorge
1015 Lausanne, Switzerland
I'm on the job market for the academic year 2025/2026!
Hey there! I’m a final-year PhD student at EPFL, where I’m advised by Matthieu Wyart and Pascal Frossard. During my PhD, I also interned as an applied scientist at Amazon’s AWS AI Labs, working in Stefano Soatto’s group.
My research is centered on understanding the fundamental principles that underpin the success of AI. In particular, my current work focuses on:
- Studying how learnability is affected by geometric properties of training data, such as symmetries, hierarchical, and compositional structures that are prevalent in images and language.
- Post-training and multi-modal aspects of large pre-trained models, particularly understanding and advancing model editing, model merging, and alignment techniques.
These directions are unified by the need to characterize the internal representations of data built by neural networks, for instance, by studying the relation between the parameter and functional spaces of these models.
Prior to my PhD, I completed a joint Master’s degree in theoretical physics from Sorbonne Université, Politecnico di Torino, SISSA, and ICTP.
selected publications
- CVPRMulti-modal hallucination control by visual information groundingIEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024
- NeurIPSLocality defeats the curse of dimensionality in convolutional teacher-student scenariosAdvances in Neural Information Processing Systems, 2021