Guardado en:
Detalles Bibliográficos
Autores principales: Zimin, Aleksandr, Polyanskiy, Yury, Rigollet, Philippe
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:https://arxiv.org/abs/2601.23236
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910041391497216
author Zimin, Aleksandr
Polyanskiy, Yury
Rigollet, Philippe
author_facet Zimin, Aleksandr
Polyanskiy, Yury
Rigollet, Philippe
contents We propose a variational framework that interprets transformer layers as iterations of an optimization algorithm acting on token embeddings. In this view, self-attention implements a gradient step of an interaction energy, while MLP layers correspond to gradient updates of a potential energy. Standard GPT-style transformers emerge as vanilla gradient descent on the resulting composite objective, implemented via Lie--Trotter splitting between these two energy functionals. This perspective enables principled architectural design using classical optimization ideas. As a proof of concept, we introduce a Nesterov-style accelerated transformer that preserves the same attention and MLP oracles. The resulting architecture consistently outperforms a nanoGPT baseline on TinyStories and OpenWebText, demonstrating that optimization-theoretic insights can translate into practical gains.
format Preprint
id arxiv_https___arxiv_org_abs_2601_23236
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle YuriiFormer: A Suite of Nesterov-Accelerated Transformers
Zimin, Aleksandr
Polyanskiy, Yury
Rigollet, Philippe
Machine Learning
Artificial Intelligence
Optimization and Control
We propose a variational framework that interprets transformer layers as iterations of an optimization algorithm acting on token embeddings. In this view, self-attention implements a gradient step of an interaction energy, while MLP layers correspond to gradient updates of a potential energy. Standard GPT-style transformers emerge as vanilla gradient descent on the resulting composite objective, implemented via Lie--Trotter splitting between these two energy functionals. This perspective enables principled architectural design using classical optimization ideas. As a proof of concept, we introduce a Nesterov-style accelerated transformer that preserves the same attention and MLP oracles. The resulting architecture consistently outperforms a nanoGPT baseline on TinyStories and OpenWebText, demonstrating that optimization-theoretic insights can translate into practical gains.
title YuriiFormer: A Suite of Nesterov-Accelerated Transformers
topic Machine Learning
Artificial Intelligence
Optimization and Control
url https://arxiv.org/abs/2601.23236