ProbMoE: Differentiable Probabilistic Routing for Mixture-of-Experts

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhao, Heng, Shao, Zilei, Broeck, Guy Van den, Zeng, Zhe
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916072119074816
author Zhao, Heng
Shao, Zilei
Broeck, Guy Van den
Zeng, Zhe
author_facet Zhao, Heng
Shao, Zilei
Broeck, Guy Van den
Zeng, Zhe
contents Mixture-of-Experts (MoE) models scale by activating only a small subset of experts per token. However, training such models remains challenging because top-$k$ routing is discrete and non-differentiable, requiring gradient estimators for expert selection whose design remains a central open problem. We introduce ProbMoE, a probabilistic routing framework that models expert selection as a distribution over cardinality-constrained expert subsets and formulates routing as probabilistic inference in this discrete subset space. We first propose ProbMoE Exact-$k$ routing, which samples $k$-expert subsets in the forward pass, and the backward pass uses gradients through each expert's exact marginal probability as a tractable surrogate for the true gradient. ProbMoE naturally generalizes to a dynamic-$k$ routing setting, where both training and inference constrain the routing cardinality to the same predefined range, allowing adaptive expert allocation per token. Across benchmarks and model backbones, ProbMoE Exact-$k$ achieves strong performance compared to competitive baselines, with improved expert utilization and routing diversity; ProbMoE Dynamic-$k$ achieves comparable performance with fewer activated experts.
format Preprint
id arxiv_https___arxiv_org_abs_2606_01509
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle ProbMoE: Differentiable Probabilistic Routing for Mixture-of-Experts
Zhao, Heng
Shao, Zilei
Broeck, Guy Van den
Zeng, Zhe
Machine Learning
Artificial Intelligence
Mixture-of-Experts (MoE) models scale by activating only a small subset of experts per token. However, training such models remains challenging because top-$k$ routing is discrete and non-differentiable, requiring gradient estimators for expert selection whose design remains a central open problem. We introduce ProbMoE, a probabilistic routing framework that models expert selection as a distribution over cardinality-constrained expert subsets and formulates routing as probabilistic inference in this discrete subset space. We first propose ProbMoE Exact-$k$ routing, which samples $k$-expert subsets in the forward pass, and the backward pass uses gradients through each expert's exact marginal probability as a tractable surrogate for the true gradient. ProbMoE naturally generalizes to a dynamic-$k$ routing setting, where both training and inference constrain the routing cardinality to the same predefined range, allowing adaptive expert allocation per token. Across benchmarks and model backbones, ProbMoE Exact-$k$ achieves strong performance compared to competitive baselines, with improved expert utilization and routing diversity; ProbMoE Dynamic-$k$ achieves comparable performance with fewer activated experts.
title ProbMoE: Differentiable Probabilistic Routing for Mixture-of-Experts
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2606.01509