Salvato in:
Dettagli Bibliografici
Autori principali: Lopez-Rubio, Ezequiel, Montes-Perez, Javier, Palomo, Esteban Jose
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:https://arxiv.org/abs/2602.05006
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866918323730513920
author Lopez-Rubio, Ezequiel
Montes-Perez, Javier
Palomo, Esteban Jose
author_facet Lopez-Rubio, Ezequiel
Montes-Perez, Javier
Palomo, Esteban Jose
contents The normalization of query and key vectors is an essential part of the Transformer architecture. It ensures that learning is stable regardless of the scale of these vectors. Some normalization approaches are available. In this preliminary work, a generalization of the QKNorm normalization scheme is proposed. The approach is based on the Lp norm, allowing non-Euclidean norms to be employed. Experimental results demonstrate the suitability of the method for a simple problem.
format Preprint
id arxiv_https___arxiv_org_abs_2602_05006
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Enhanced QKNorm normalization for neural transformers with the Lp norm
Lopez-Rubio, Ezequiel
Montes-Perez, Javier
Palomo, Esteban Jose
Machine Learning
Artificial Intelligence
Computation and Language
68T07
The normalization of query and key vectors is an essential part of the Transformer architecture. It ensures that learning is stable regardless of the scale of these vectors. Some normalization approaches are available. In this preliminary work, a generalization of the QKNorm normalization scheme is proposed. The approach is based on the Lp norm, allowing non-Euclidean norms to be employed. Experimental results demonstrate the suitability of the method for a simple problem.
title Enhanced QKNorm normalization for neural transformers with the Lp norm
topic Machine Learning
Artificial Intelligence
Computation and Language
68T07
url https://arxiv.org/abs/2602.05006