What Gets Activated: Uncovering Domain and Driver Experts in MoE Language Models

TL;DR

This paper investigates expert activation patterns in MoE language models, identifying domain and driver experts, and demonstrates how understanding these can improve model interpretability and performance across multiple domains.

Contribution

It introduces entropy-based and causal-effect metrics to identify domain and driver experts, revealing their roles and influence in MoE models, and shows how expert weight adjustments enhance performance.

Findings

Some experts show clear domain preferences.

Certain experts have a strong causal influence on performance.

Adjusting expert weights improves model performance across domains.

Abstract

Most interpretability work focuses on layer- or neuron-level mechanisms in Transformers, leaving expert-level behavior in MoE LLMs underexplored. Motivated by functional specialization in the human brain, we analyze expert activation by distinguishing domain and driver experts. In this work, we study expert activation in MoE models across three public domains and address two key questions: (1) which experts are activated, and whether certain expert types exhibit consistent activation patterns; and (2) how tokens are associated with and trigger the activation of specific experts. To answer these questions, we introduce entropy-based and causal-effect metrics to assess whether an expert is strongly favored for a particular domain, and how strongly expert activation contributes causally to the model's output, thus identify domain and driver experts, respectively. Furthermore, we explore how individual tokens are associated with the activation of specific experts. Our analysis reveals that (1) Among the activated experts, some show clear domain preferences, while others exert strong causal influence on model performance, underscoring their decisive roles. (2) tokens occurring earlier in a sentence are more likely to trigger the driver experts, and (3) adjusting the weights of domain and driver experts leads to significant performance gains across all three models and domains. These findings shed light on the internal mechanisms of MoE models and enhance their interpretability.