# Beyond Benchmarks: Understanding Mixture-of-Experts Models through Internal Mechanisms

arXiv: https://arxiv.org/abs/2509.23933 · HTML: https://mlsanity.com/paper/2509.23933/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Placed in MoE theory and gating functions, one of 12 directions in Mixture-of-experts.
- Builds on 15 of its 32 references that are in the map.
- First posted on arXiv 28 Sep 2025.

## Placed in

- Mixture-of-experts / [MoE theory and gating functions](https://mlsanity.com/moe/moe-theory-and-gating-functions/)

## Builds on (most cited first, 15 of 15)

- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [Program Synthesis with Large Language Models](https://mlsanity.com/paper/2108.07732/) (2021)
- [Mixtral of Experts](https://mlsanity.com/paper/2401.04088/) (2024)
- [gpt-oss-120b & gpt-oss-20b Model Card](https://mlsanity.com/paper/2508.10925/) (2025)
- [Transformer Feed-Forward Layers Are Key-Value Memories](https://mlsanity.com/paper/2012.14913/) (2020)
- [DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model](https://mlsanity.com/paper/2405.04434/) (2024)
- [DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence](https://mlsanity.com/paper/2406.11931/) (2024)
- [OLMoE: Open Mixture-of-Experts Language Models](https://mlsanity.com/paper/2409.02060/) (2024)
- [Auxiliary-Loss-Free Load Balancing Strategy for Mixture-of-Experts](https://mlsanity.com/paper/2408.15664/) (2024)
- [Don't Make Your LLM an Evaluation Benchmark Cheater](https://mlsanity.com/paper/2311.01964/) (2023)
- [Model Utility Law: Evaluating LLMs beyond Performance through Mechanism Interpretable Metric](https://mlsanity.com/paper/2504.07440/) (2025)
- [EffiEval: Efficient and Generalizable Model Evaluation via Capability Coverage Maximization](https://mlsanity.com/paper/2508.09662/) (2025)

## Built on by (most cited first, 7 of 7)

- [XShare: Collaborative in-Batch Expert Sharing for Faster MoE Inference](https://mlsanity.com/paper/2602.07265/) (2026)
- [Knowledge Localization in Mixture-of-Experts LLMs Using Cross-Lingual Inconsistency](https://mlsanity.com/paper/2603.17102/) (2026)
- [Does the Same Token Mean the Same State? MoE Routing as Signal for Reasoning Control](https://mlsanity.com/paper/2606.22798/) (2026)
- [DBES: A Systematic Benchmark and Metric Suite for Evaluating Expert Specialization in Large-Scale MoEs](https://mlsanity.com/paper/2605.18498/) (2026)
- [Beyond Routing: Characterising Expert Tuning and Representation in Vision Mixture-of-Experts](https://mlsanity.com/paper/2605.20610/) (2026)
- [From Concentration to Differentiation and Back: Routing Effective Rank in MoE Reasoning Cohorts](https://mlsanity.com/paper/2609.06403/) (2026)
- [Belief-Trajectory Energy: Measuring the Path to a Prediction](https://mlsanity.com/paper/2610.05114/) (2026)

## Abstract (arXiv)

Mixture-of-Experts (MoE) architectures have emerged as a promising direction, offering efficiency and scalability by activating only a subset of parameters during inference. However, current research remains largely performance-centric, with limited understanding of its internal mechanisms, thereby constraining broader progress. In this work, we use an internal metric to investigate the mechanisms of MoE architecture by explicitly incorporating routing mechanisms and analyzing expert-level behaviors. Through systematic analyses of a wide range of publicly available MoE models, we uncover several findings: (1) neuron utilization decreases as models evolve, reflecting stronger generalization; (2) training exhibits a dynamic trajectory, where benchmark performance alone provides limited signal while MUI reveals deeper insights; (3) task completion emerges from collaborative contributions of multiple experts, with shared experts driving concentration; and (4) activation patterns at the neuron level provide a fine-grained proxy for data diversity. Together, these results demonstrate the potential of MUI as a complementary indicator to benchmark performance, offering new insights into the capacity, dynamics, and specialization of MoE models. Our project can be found at https://yingjiahao14.github.io/MoE-MUI/.