A multiscale analysis of mean-field transformers in the moderate interaction regime

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Bruno, Giuseppe, Pasqualotto, Federico, Agazzi, Andrea
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909814851895296
author Bruno, Giuseppe
Pasqualotto, Federico
Agazzi, Andrea
author_facet Bruno, Giuseppe
Pasqualotto, Federico
Agazzi, Andrea
contents In this paper, we study the evolution of tokens through the depth of encoder-only transformer models at inference time by modeling them as a system of particles interacting in a mean-field way and studying the corresponding dynamics. More specifically, we consider this problem in the moderate interaction regime, where the number $N$ of tokens is large and the inverse temperature parameter $β$ of the model scales together with $N$. In this regime, the dynamics of the system displays a multiscale behavior: a fast phase, where the token empirical measure collapses on a low-dimensional space, an intermediate phase, where the measure further collapses into clusters, and a slow one, where such clusters sequentially merge into a single one. We provide a rigorous characterization of the limiting dynamics in each of these phases and prove convergence in the above mentioned limit, exemplifying our results with some simulations.
format Preprint
id arxiv_https___arxiv_org_abs_2509_25040
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A multiscale analysis of mean-field transformers in the moderate interaction regime
Bruno, Giuseppe
Pasqualotto, Federico
Agazzi, Andrea
Machine Learning
Probability
In this paper, we study the evolution of tokens through the depth of encoder-only transformer models at inference time by modeling them as a system of particles interacting in a mean-field way and studying the corresponding dynamics. More specifically, we consider this problem in the moderate interaction regime, where the number $N$ of tokens is large and the inverse temperature parameter $β$ of the model scales together with $N$. In this regime, the dynamics of the system displays a multiscale behavior: a fast phase, where the token empirical measure collapses on a low-dimensional space, an intermediate phase, where the measure further collapses into clusters, and a slow one, where such clusters sequentially merge into a single one. We provide a rigorous characterization of the limiting dynamics in each of these phases and prove convergence in the above mentioned limit, exemplifying our results with some simulations.
title A multiscale analysis of mean-field transformers in the moderate interaction regime
topic Machine Learning
Probability
url https://arxiv.org/abs/2509.25040