# Routing Sensitivity Without Controllability: A Diagnostic Study of Fairness in MoE Language Models

arXiv: https://arxiv.org/abs/2603.27141 · HTML: https://mlsanity.com/paper/2603.27141/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Expert routing strategies, one of 12 directions in Mixture-of-experts.
- Builds on 22 of its 26 references that are in the map.
- First posted on arXiv 28 Mar 2026.

## Placed in

- Mixture-of-experts / [Expert routing strategies](https://mlsanity.com/moe/expert-routing-strategies/)

## Builds on (most cited first, 22 of 22)

- [Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer](https://mlsanity.com/paper/1701.06538/) (2017)
- [Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity](https://mlsanity.com/paper/2101.03961/) (2021)
- [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://mlsanity.com/paper/2109.07958/) (2021)
- [GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding](https://mlsanity.com/paper/2006.16668/) (2020)
- [Language (Technology) is Power: A Critical Survey of "Bias" in NLP](https://mlsanity.com/paper/2005.14050/) (2020)
- [StereoSet: Measuring stereotypical bias in pretrained language models](https://mlsanity.com/paper/2004.09456/) (2020)
- [CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Models](https://mlsanity.com/paper/2010.00133/) (2020)
- [BBQ: A Hand-Built Bias Benchmark for Question Answering](https://mlsanity.com/paper/2110.08193/) (2021)
- [BOLD: Dataset and Metrics for Measuring Biases in Open-Ended Language Generation](https://mlsanity.com/paper/2101.11718/) (2021)
- [Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLP](https://mlsanity.com/paper/2103.00453/) (2021)
- [OLMoE: Open Mixture-of-Experts Language Models](https://mlsanity.com/paper/2409.02060/) (2024)
- [Counterfactual Data Augmentation for Mitigating Gender Stereotypes in Languages with Rich Morphology](https://mlsanity.com/paper/1906.04571/) (2019)
- [MaxMin-RLHF: Alignment with Diverse Human Preferences](https://mlsanity.com/paper/2402.08925/) (2024)
- [Self-Debiasing Large Language Models: Zero-Shot Recognition and Reduction of Stereotypes](https://mlsanity.com/paper/2402.01981/) (2024)
- [FairSteer: Inference Time Debiasing for LLMs with Dynamic Activation Steering](https://mlsanity.com/paper/2504.14492/) (2025)
- [Prompting Techniques for Reducing Social Bias in LLMs through System 1 and System 2 Cognitive Processes](https://mlsanity.com/paper/2404.17218/) (2024)
- [Towards an empirical understanding of MoE design choices](https://mlsanity.com/paper/2402.13089/) (2024)
- [SAFEx: Analyzing Vulnerabilities of MoE-Based LLMs via Stable Safety-critical Expert Identification](https://mlsanity.com/paper/2506.17368/) (2025)
- [Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training](https://mlsanity.com/paper/2505.14681/) (2025)
- [Shifting Perspectives: Steering Vectors for Robust Bias Mitigation in LLMs](https://mlsanity.com/paper/2503.05371/) (2025)
- [R2-T2: Re-Routing in Test-Time for Multimodal Mixture-of-Experts](https://mlsanity.com/paper/2502.20395/) (2025)
- [MoE Lens -- An Expert Is All You Need](https://mlsanity.com/paper/2603.05806/) (2026)

## Built on by (most cited first, 1 of 1)

- [THESIS-MoE: Trainable Hierarchical Extraction and SteerIng of Sycophancy in Mixture-of-Experts](https://mlsanity.com/paper/2608.15687/) (2026)

## Abstract (arXiv)

Demographic prompts are routed differently from neutral prompts in Mixture-of-Experts (MoE) language models, motivating tests of routing-level stereotype control. We introduce Fairness-Aware Routing Equilibrium (FARE), a diagnostic framework combining demographic routing profiles, empirical layer selection, and fixed inference-time reweighting, and evaluate five MoE architectures in English. At the selected operating points, CrowS-Pairs preference changes by at most 1.3 percentage points; DeepSeekMoE selects no intervention. Paired 95% confidence intervals exclude decreases larger than 2.2 points on each intervened model, and the only nominally significant change (Qwen1.5, p = 0.015) does not survive multiple-comparison correction. OLMoE and Qwen3 nevertheless change nearly every top-k expert set. Noise controls, random and truncated synthetic profiles, and hard masking also move preference by at most 1.5 points. Four generation protocols on OLMoE, Mixtral, Qwen1.5, and Qwen3 show no consistent change in the measured toxicity, lexical, or reference-overlap metrics. Selection and evaluation items overlap, so these comparisons are not independent evaluations. The tested reweighting procedure offers limited stereotype control.