Does Role Specialization Matter for Explanation Faithfulness in Mixture-of-Experts?

2026-06-28

AI Summary

This work addresses the limited faithfulness of explanations in existing Mixture-of-Experts (MoE) models, which stems primarily from high representational overlap among experts that undermines role specialization. To mitigate this issue, the authors propose a representation-level decorrelation regularization mechanism that explicitly reduces the similarity between experts’ latent representations, thereby enhancing their functional distinctiveness. The method consistently improves key faithfulness metrics—including comprehensiveness, sufficiency, and AOPC—across multiple multimodal benchmarks without compromising task performance. Furthermore, the approach demonstrates broad applicability and effectiveness within standard sparse MoE architectures, validating its potential as a general-purpose enhancement for interpretable expert-based models.

Abstract

Mixture-of-Experts (MoE) architectures have recently been extended with role-based mechanisms for interpretability. This is typically done by assigning semantic roles to individual expert components, for example roles like synergy, redundancy, and uniqueness in multimodal settings. However, whether such structural role decomposition preserves explanation faithfulness of the overall architecture remains largely underexplored. We hypothesize that inter-expert representation overlap weakens effective role separation and degrades attribution-based faithfulness, even when semantic roles are explicitly defined. To address this limitation, we introduce representation-level decorrelation regularization to explicitly reduce inter-expert similarity in latent space. Using representation decorrelation objectives, we encourage clearer specialization among experts by minimizing representation overlap. Our experiments show that across multiple multimodal benchmarks, this separation consistently improves explanation faithfulness, as measured by comprehensiveness, sufficiency, and their Area Over the Perturbation Curve (AOPC) summaries, while preserving task performance. We further show that these improvements are not limited to role-based architectures such as Interpretable Multimodal Interaction-aware MoE (I2MoE). Similar trends are observed in a standard sparse MoE baseline, suggesting that representation-level separation may provide a more general mechanism for enhancing explanation faithfulness in MoE systems. Overall, our findings suggest that structural role decomposition alone may be insufficient to guarantee faithful explanations and that representation-level separation helps improve explanation faithfulness. To support reproducibility, the source code and supplementary material are publicly available at https://github.com/dut0817/FL-I2MoE_Decor.

Problem

• Mixture-of-Experts
• Explanation Faithfulness
• Role Specialization
• Representation Overlap
• Interpretability

Innovation

• Mixture-of-Experts
• explanation faithfulness
• representation decorrelation
• role specialization
• interpretability