Talks and Posters
2026
Talks
Tutorial: how to build and train a GPT 🔽 Slides - Open Notebook
Statistical Physics of Deep Learning III - Lake Como School of Advanced studies, 16/06/2026
Implicit bias produces scaling laws in learning curves, from perceptrons to deep networks 🔽 Slides
PCSL group meeting - École polytechnique fédérale de Lausanne (EPFL), 06/03/2026
ICLR26 short recorded talk - Video
Pseudolikelihood produces associative memories able to generalize, even for asymmetric couplings 🔽 Slides
4th workshop of Unione Matematica Italiana (UMI) - Sapienza University of Rome, 23/01/2026
Posters
Implicit bias produces neural scaling laws in learning curves, from perceptrons to deep networks 🔽 Poster
ICLR2026 - Rio de Janeiro, 23-27/04/2026
Statistical Physics of Deep Learning III - Lake Como School of Advanced studies, 15-19/06/2026
Power-law feature statistics explain test reconstruction gaps in Associative Memories 🔽 Poster
NFAM2026 - Rio de Janeiro, 23-27/04/2026
2025
Talks
Scaling laws, from Perceptrons to Deep Networks 🔽 Slides
Sapienza University of Rome - Chimera group Journal Club, 07/10/2025
University of Parma - Statistical mechanics group, 23/10/2025
Universidad Complutense de Madrid - Theoretical physics department seminar, 18/02/2026
Statistical mechanics of autoregressive models: towards a theory of Self-Attention 🔽 Slides
PhD talks - Sapienza University of Rome, 16/10/2025
Posters
Pseudo-likelihood produces associative memories able to generalize, even for asymmetric couplings 🔽 Poster
Mathematical Methods for High-Dimensional Data, Sapienza University of Rome, 8-12/09/25
STATPHYS29, Firenze, 13-17/07/25
Implicit bias produces neural scaling laws in learning curves, from perceptrons to deep networks 🔽 Poster
ROccella Conference on Inference and AI, Roccella Ionica, 1-5/09/25
Learning and Optimization in High Dimensions, Bocconi Univeristy, Milano, 8-11/07/25
Beg Rohu summer school, Saint-Pierre-Quiberon, 2-14/06/25
2024
Talks
Self-attention as an attractor network: transient memories without backpropagation 🔽 Slides
SISSA Journal Club — Trieste, 11/10/2024
Posters
Self-attention as an attractor network: transient memories without backpropagation 🔽 Poster
ROccella Conference on Inference and AI, Roccella Ionica, 1-5/09/24
SISSA Journal Club — Trieste, 11/10/2024