A Minimal Bifurcation Model of Load Imbalance in a Softmax Mixture-of-Experts Router

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteur principal: Kiselev, O. M.
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866917542423953408
author Kiselev, O. M.
author_facet Kiselev, O. M.
contents We propose a minimal dynamical model of adaptive softmax routing for a two-expert Mixture-of-Experts (MoE) layer. The model is obtained as a mean-field limit of a discrete reinforcement rule: the selected expert receives a small score increment, while all scores undergo regularizing decay. In the symmetric case the limiting system has a supercritical pitchfork bifurcation: for weak feedback there is a unique stable balanced state, whereas above a critical feedback strength two stable asymmetric states appear. When an external asymmetry is added, the pitchfork unfolds into a pair of fold bifurcations forming a cusp in the control-parameter plane. We derive exact parametric equations for the bifurcation set and the local normal form of the cusp catastrophe. Numerical experiments connect this picture to empirical expert load, a small trainable MoE model, hard top-1 PyTorch routing, and a small classification experiment on digits. The results provide a controlled low-dimensional mechanism for abrupt transitions to load imbalance in adaptive MoE routers.
format Preprint
id arxiv_https___arxiv_org_abs_2605_29121
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle A Minimal Bifurcation Model of Load Imbalance in a Softmax Mixture-of-Experts Router
Kiselev, O. M.
Dynamical Systems
Artificial Intelligence
Machine Learning
We propose a minimal dynamical model of adaptive softmax routing for a two-expert Mixture-of-Experts (MoE) layer. The model is obtained as a mean-field limit of a discrete reinforcement rule: the selected expert receives a small score increment, while all scores undergo regularizing decay. In the symmetric case the limiting system has a supercritical pitchfork bifurcation: for weak feedback there is a unique stable balanced state, whereas above a critical feedback strength two stable asymmetric states appear. When an external asymmetry is added, the pitchfork unfolds into a pair of fold bifurcations forming a cusp in the control-parameter plane. We derive exact parametric equations for the bifurcation set and the local normal form of the cusp catastrophe. Numerical experiments connect this picture to empirical expert load, a small trainable MoE model, hard top-1 PyTorch routing, and a small classification experiment on digits. The results provide a controlled low-dimensional mechanism for abrupt transitions to load imbalance in adaptive MoE routers.
title A Minimal Bifurcation Model of Load Imbalance in a Softmax Mixture-of-Experts Router
topic Dynamical Systems
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2605.29121