Saved in:
Bibliographic Details
Main Author: Nunley, Joshua
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.18417
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910028231868416
author Nunley, Joshua
author_facet Nunley, Joshua
contents This paper presents a direct framework for sequence models with hidden states on closed subgroups of U(d). We use a minimal axiomatic setup and derive recurrent and transformer templates from a shared skeleton in which subgroup choice acts as a drop-in replacement for state space, tangent projection, and update map. We then specialize to O(d) and evaluate orthogonal-state RNN and transformer models on Tiny Shakespeare and Penn Treebank under parameter-matched settings. We also report a general linear-mixing extension in tangent space, which applies across subgroup choices and improves finite-budget performance in the current O(d) experiments.
format Preprint
id arxiv_https___arxiv_org_abs_2602_18417
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Subgroups of $U(d)$ Induce Natural RNN and Transformer Architectures
Nunley, Joshua
Machine Learning
Computation and Language
68T07, 22E70
I.2.6; G.3
This paper presents a direct framework for sequence models with hidden states on closed subgroups of U(d). We use a minimal axiomatic setup and derive recurrent and transformer templates from a shared skeleton in which subgroup choice acts as a drop-in replacement for state space, tangent projection, and update map. We then specialize to O(d) and evaluate orthogonal-state RNN and transformer models on Tiny Shakespeare and Penn Treebank under parameter-matched settings. We also report a general linear-mixing extension in tangent space, which applies across subgroup choices and improves finite-budget performance in the current O(d) experiments.
title Subgroups of $U(d)$ Induce Natural RNN and Transformer Architectures
topic Machine Learning
Computation and Language
68T07, 22E70
I.2.6; G.3
url https://arxiv.org/abs/2602.18417