Cover Image

Single-Head Attention in High Dimensions: A Theory of Generalization, Weights Spectra, and Scaling Laws

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Boncoraglio, Fabrizio, Erba, Vittorio, Troiani, Emanuele, Xu, Yizhou, Krzakala, Florent, Zdeborová, Lenka
Format:	Preprint
Published:	2025
Subjects:	Machine Learning Disordered Systems and Neural Networks Information Theory
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Bayes optimal learning of attention-indexed models
by: Boncoraglio, Fabrizio, et al.
Published: (2025)

Scaling Laws and Spectra of Shallow Neural Networks in the Feature Learning Regime
by: Defilippis, Leonardo, et al.
Published: (2025)

The Nuclear Route: Sharp Asymptotics of ERM in Overparameterized Quadratic Networks
by: Erba, Vittorio, et al.
Published: (2025)

Learning with Restricted Boltzmann Machines: Asymptotics of AMP and GD in High Dimensions
by: Xu, Yizhou, et al.
Published: (2025)

Quenches in the Sherrington-Kirkpatrick model
by: Erba, Vittorio, et al.
Published: (2024)

Statistical mechanics of the maximum-average submatrix problem
by: Erba, Vittorio, et al.
Published: (2023)

Fundamental limits of learning in sequence multi-index models and deep attention networks: High-dimensional asymptotics and sharp thresholds
by: Troiani, Emanuele, et al.
Published: (2025)

Bilinear Sequence Regression: A Model for Learning from Long Sequences of High-dimensional Tokens
by: Erba, Vittorio, et al.
Published: (2024)

The Rules-and-Facts Model for Simultaneous Generalization and Memorization in Neural Networks
by: Farné, Gabriele, et al.
Published: (2026)

Bayes-optimal learning of an extensive-width neural network from quadratically many samples
by: Maillard, Antoine, et al.
Published: (2024)

Asymptotics of SGD in Sequence-Single Index Models and Single-Layer Attention Networks
by: Arnaboldi, Luca, et al.
Published: (2025)

Fundamental Limits of Matrix Sensing: Exact Asymptotics, Universality, and Applications
by: Xu, Yizhou, et al.
Published: (2025)

Fundamental computational limits of weak learnability in high-dimensional multi-index models
by: Troiani, Emanuele, et al.
Published: (2024)

Inference in Spreading Processes with Neural-Network Priors
by: Ghio, Davide, et al.
Published: (2025)

Computational Thresholds in Multi-Modal Learning via the Spiked Matrix-Tensor Model
by: Tabanelli, Hugo, et al.
Published: (2025)

On the Atypical Solutions of the Symmetric Binary Perceptron
by: Barbier, Damien, et al.
Published: (2023)

A solvable high-dimensional model where nonlinear autoencoders learn structure invisible to PCA while test loss misaligns with generalization
by: Mendes, Vicente Conde, et al.
Published: (2026)

Low-rank Matrix Estimation with Inhomogeneous Noise
by: Guionnet, Alice, et al.
Published: (2022)

Sampling with flows, diffusion and autoregressive neural networks: A spin-glass perspective
by: Ghio, Davide, et al.
Published: (2023)

The phase diagram of compressed sensing with $\ell_0$-norm regularization
by: Barbier, Damien, et al.
Published: (2024)

Analysis of Bootstrap and Subsampling in High-dimensional Regularized Regression
by: Clarté, Lucas, et al.
Published: (2024)

The maximum-average subtensor problem: equilibrium and out-of-equilibrium properties
by: Erba, Vittorio, et al.
Published: (2025)

Gaussian Universality of Perceptrons with Random Labels
by: Gerace, Federica, et al.
Published: (2022)

High-dimensional Asymptotics of Denoising Autoencoders
by: Cui, Hugo, et al.
Published: (2023)

Asymptotics of feature learning in two-layer networks after one gradient-step
by: Cui, Hugo, et al.
Published: (2024)

The committee machine: Computational to statistical gaps in learning a two-layers neural network
by: Aubin, Benjamin, et al.
Published: (2018)

Building Conformal Prediction Intervals with Approximate Message Passing
by: Clarté, Lucas, et al.
Published: (2024)

Optimal thresholds and algorithms for a model of multi-modal learning in high dimensions
by: Keup, Christian, et al.
Published: (2024)

Dynamical Cavity Method for Hypergraphs and its Application to Quenches in the k-XOR-SAT Problem
by: Maier, Aude, et al.
Published: (2024)

Dynamical Phase Transitions in Graph Cellular Automata
by: Behrens, Freya, et al.
Published: (2023)

Minority Takeover in Majority Dynamics: Searching for Rare Initializations via the History Passing Algorithm
by: Jankola, Marek, et al.
Published: (2025)

Spectral Thresholds in Correlated Spiked Models and Fundamental Limits of Partial Least Squares
by: Mergny, Pierre, et al.
Published: (2025)

Counting and Hardness-of-Finding Fixed Points in Cellular Automata on Random Graphs
by: Koller, Cédric, et al.
Published: (2024)

A High Dimensional Statistical Model for Adversarial Training: Geometry and Trade-Offs
by: Tanner, Kasimir, et al.
Published: (2024)

Deep Learning as Neural Low-Degree Filtering: A Spectral Theory of Hierarchical Feature Learning
by: Dandi, Yatin, et al.
Published: (2026)

Optimal Spectral Transitions in High-Dimensional Multi-Index Models
by: Defilippis, Leonardo, et al.
Published: (2025)

On the existence of consistent adversarial attacks in high-dimensional linear classification
by: Vilucchio, Matteo, et al.
Published: (2025)

Statistical Advantage of Softmax Attention: Insights from Single-Location Regression
by: Duranthon, O., et al.
Published: (2025)

Rigorous Asymptotics for First-Order Algorithms Through the Dynamical Cavity Method
by: Dandi, Yatin, et al.
Published: (2026)

Sequential Dynamics in Ising Spin Glasses
by: Dandi, Yatin, et al.
Published: (2025)