On the Spatial Structure of Mixture-of-Experts in Transformers

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Bershatsky, Daniel, Oseledets, Ivan
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866908304291135488
author Bershatsky, Daniel
Oseledets, Ivan
author_facet Bershatsky, Daniel
Oseledets, Ivan
contents A common assumption is that MoE routers primarily leverage semantic features for expert selection. However, our study challenges this notion by demonstrating that positional token information also plays a crucial role in routing decisions. Through extensive empirical analysis, we provide evidence supporting this hypothesis, develop a phenomenological explanation of the observed behavior, and discuss practical implications for MoE-based architectures.
format Preprint
id arxiv_https___arxiv_org_abs_2504_04444
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle On the Spatial Structure of Mixture-of-Experts in Transformers
Bershatsky, Daniel
Oseledets, Ivan
Computation and Language
Artificial Intelligence
Machine Learning
A common assumption is that MoE routers primarily leverage semantic features for expert selection. However, our study challenges this notion by demonstrating that positional token information also plays a crucial role in routing decisions. Through extensive empirical analysis, we provide evidence supporting this hypothesis, develop a phenomenological explanation of the observed behavior, and discuss practical implications for MoE-based architectures.
title On the Spatial Structure of Mixture-of-Experts in Transformers
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2504.04444