Salvato in:
Dettagli Bibliografici
Autore principale: Daneshmand, Hadi
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2410.19931
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909967928262656
author Daneshmand, Hadi
author_facet Daneshmand, Hadi
contents Despite their empirical success, the internal mechanism by which transformer models align tokens during language processing remains poorly understood. This paper provides a mechanistic and theoretical explanation of token alignment in LLMs. We first present empirical evidences showing that, in machine translation, attention weights progressively align translated word pairs across layers, closely approximating Optimal Transport (OT) between word embeddings. Building on this observation, we prove that softmax self-attention layers can simulate gradient descent on the dual of the entropy-regularized OT problem, providing a theoretical foundation for the alignment. Our analysis yields a constructive convergence bound showing that transformer depth controls OT approximation accuracy. A direct implication is that standard transformers can sort lists of varying lengths without any parameter adjustment, up to an error term vanishing with transformers depth.
format Preprint
id arxiv_https___arxiv_org_abs_2410_19931
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Provable optimal transport with transformers: The essence of depth and prompt engineering
Daneshmand, Hadi
Machine Learning
Optimization and Control
Despite their empirical success, the internal mechanism by which transformer models align tokens during language processing remains poorly understood. This paper provides a mechanistic and theoretical explanation of token alignment in LLMs. We first present empirical evidences showing that, in machine translation, attention weights progressively align translated word pairs across layers, closely approximating Optimal Transport (OT) between word embeddings. Building on this observation, we prove that softmax self-attention layers can simulate gradient descent on the dual of the entropy-regularized OT problem, providing a theoretical foundation for the alignment. Our analysis yields a constructive convergence bound showing that transformer depth controls OT approximation accuracy. A direct implication is that standard transformers can sort lists of varying lengths without any parameter adjustment, up to an error term vanishing with transformers depth.
title Provable optimal transport with transformers: The essence of depth and prompt engineering
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2410.19931