Salvato in:
Dettagli Bibliografici
Autore principale: Racioppo, Peter
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:https://arxiv.org/abs/2605.11007
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917481489104896
author Racioppo, Peter
author_facet Racioppo, Peter
contents We show that the core components of the Transformer block -- attention, residual connections, and normalization -- arise naturally from a single geometric estimation problem. Modeling the latent state as a direction on the hypersphere, with noise defined in the tangent plane at the current estimate, yields a precision-weighted directional inference procedure in which attention aggregates evidence, residual connections implement incremental state updates, and normalization retracts the updated state back onto the hypersphere. Together, these components follow from the geometry of the estimation problem rather than being introduced as independent architectural choices.
format Preprint
id arxiv_https___arxiv_org_abs_2605_11007
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle RT-Transformer: The Transformer Block as a Spherical State Estimator
Racioppo, Peter
Machine Learning
Artificial Intelligence
We show that the core components of the Transformer block -- attention, residual connections, and normalization -- arise naturally from a single geometric estimation problem. Modeling the latent state as a direction on the hypersphere, with noise defined in the tangent plane at the current estimate, yields a precision-weighted directional inference procedure in which attention aggregates evidence, residual connections implement incremental state updates, and normalization retracts the updated state back onto the hypersphere. Together, these components follow from the geometry of the estimation problem rather than being introduced as independent architectural choices.
title RT-Transformer: The Transformer Block as a Spherical State Estimator
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2605.11007