0

Autonomy-of-Experts Models

Autonomy-of-Experts (AoE) improves MoE models by enabling experts to autonomously select inputs based on self-evaluated capacity, reducing the need for routers and enhancing performance.

Year
2025
Venue
arXiv 2025
Authors
8
Hosting
Abstract onlyARXIV-DEFAULT

Cite

Notes

Only stored in your browser.

Attribution

Abstract & full text
arxiv.org/abs/2501.13074ARXIV-DEFAULT
TL;DR
Semantic Scholar
Attribution policy →

Abstract

Mixture-of-Experts (MoE) models mostly use a router to assign tokens to specific expert modules, activating only partial parameters and often outperforming dense models. We argue that the separation between the router's decision-making and the experts' execution is a critical yet overlooked issue, leading to suboptimal expert selection and ineffective learning. To address this, we propose Autonomy-of-Experts (AoE), a novel MoE paradigm in which experts autonomously select themselves to process inputs. AoE is based on the insight that an expert is aware of its own capacity to effectively process a token, an awareness reflected in the scale of its internal activations. In AoE, routers are removed; instead, experts pre-compute internal activations for inputs and are ranked based on their activation norms. Only the top-ranking experts proceed with the forward pass, while the others abort. The overhead of pre-computing activations is reduced through a low-rank weight factorization. This self-evaluating-then-partner-comparing approach ensures improved expert selection and effective learning. We pre-train language models having 700M up to 4B parameters, demonstrating that AoE outperforms traditional MoE models with comparable efficiency.

Authors

8