When Are Experts Misrouted? Counterfactual Routing Analysis in Mixture-of-Experts Language Models

• URL: http://arxiv.org/abs/2605.07260v1
• Date: Fri, 08 May 2026 05:26:09 GMT
• Title: When Are Experts Misrouted? Counterfactual Routing Analysis in Mixture-of-Experts Language Models
• Abstract summary: Mixture-of-Experts (MoE) language models route each token to a small subset of experts.<n>We compare each standard route against sampled equal-compute alternatives for the same token and score each by the next-token probability it assigns to the realized token in a verified reasoning trajectory.<n>The result is sharply token-conditional: the standard router is well-aligned with route utility on confident tokens but uninformative on the fragile tokens that drive hard reasoning.
• Score: 27.942734943134983
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Mixture-of-Experts (MoE) language models route each token to a small subset of experts, but whether the routes selected by a trained top-$k$ router are good ones is rarely evaluated directly. Holding the model fixed, we compare each standard route against sampled equal-compute alternatives for the same token and score each by the next-token probability it assigns to the realized token in a verified reasoning trajectory. The result is sharply token-conditional: the standard router is well-aligned with route utility on confident tokens but uninformative on the fragile tokens that drive hard reasoning, where lower-loss equal-compute routes consistently exist inside the frozen model but are not selected. The same pattern holds across Qwen3-30B-A3B, GPT-OSS-20B, DeepSeek-V2-Lite, and OLMoE-1B-7B, and follows structurally from how standard top-$k$ training evaluates routing decisions: the language modeling loss scores only the executed route, and load balancing depends only on aggregate routing statistics. A minimal router-only update to the final-layer router, leaving every expert and every other router frozen, is sufficient to shift pass@K on AIME 2024+2025 and HMMT 2025 for both Qwen3-30B-A3B and GPT-OSS-20B, suggesting that at least part of the failure reflects router-reachable misallocation rather than expert capacity alone.

Related papers

• Agent-as-a-Router: Agentic Model Routing for Coding Tasks [54.00153674547733]
  Agent-as-a-router (AC) formalizes routing as a C-A-F loop (Context->Action->FeedbackContext)<n>AC achieves the lowest cumulative regret on in-distribution tasks and generalizes to out-of-distribution agentic-programming tasks.
  arXiv  Detail & Related papers  (2026-06-22T06:37:31Z)
• TwinRouterBench: Fast Static and Live Dynamic Evaluation for Realistic Agentic LLM Routing [24.63595849324157]
  We introduce TwinBench, a step-level routing benchmark with two tracks.<n>The static track provides 970 router-visible prefixes from 520 instances across SWE-bench, BFCL, mtRAG, QMSum, and PinchBench.<n>The dynamic track supplies a harness that runs routers on the full 500-case SWE-bench Verified suite.
  arXiv  Detail & Related papers  (2026-05-14T08:58:59Z)
• LatentRouter: Can We Choose the Right Multimodal Model Before Seeing Its Answer? [69.71754384259167]
  Multimodal large language models (MLLMs) have heterogeneous strengths across OCR, chart understanding, spatial reasoning, visual question answering, cost, and latency.<n>We propose Latent, a router that formulates MLLM routing as counterfactual multimodal utility prediction.<n> Experiments on MMR-Bench and VL-Bench show that Latent outperforms fixed-model, feature-level, and learned-router baselines.
  arXiv  Detail & Related papers  (2026-05-11T22:42:12Z)
• Adaptive Inverted-Index Routing for Granular Mixtures-of-Experts [49.09151538536423]
  Mixture-of-experts (MoE) models enable scalable transformer architectures by activating only a subset of experts per token.<n>Recent evidence suggests that performance improves with increasingly granular experts, i.e., many small experts instead of a few large ones.<n>We introduce adaptive inverted-index routing for MoE (AIR-MoE), an inverted-index-inspired routing architecture based on vector quantization (VQ)
  arXiv  Detail & Related papers  (2026-05-06T14:15:10Z)
• Self-Routing: Parameter-Free Expert Routing from Hidden States [14.018297453424454]
  Mixture-of-Experts (MoE) layers increase model capacity by activating only a small subset of experts per token.<n>We propose Self-Routing, a parameter-free routing mechanism that uses a designated subspace of the token hidden state directly as expert logits.
  arXiv  Detail & Related papers  (2026-04-01T03:05:20Z)
• When Routing Collapses: On the Degenerate Convergence of LLM Routers [46.01380774114097]
  As user's cost budget increases, routers systematically default to the most capable and most expensive model.<n>We propose Equi, a decision-aware router that directly learns model rankings.<n>On RouterBench, Equi reduces cost by about 17% at GPT-4-level performance compared to the strongest prior router.
  arXiv  Detail & Related papers  (2026-02-03T12:51:55Z)
• Routing by Analogy: kNN-Augmented Expert Assignment for Mixture-of-Experts [32.65737144630759]
  Mixture-of-Experts (MoE) architectures scale large language models efficiently by employing a parametric "router" to dispatch tokens to a sparse subset of experts.<n>We introduce kNN-MoE, a retrieval-augmented routing framework that reuses optimal expert assignments from a memory of similar past cases.<n>Experiments show kNN-MoE outperforms zero-shot baselines and rivals computationally expensive supervised fine-tuning.
  arXiv  Detail & Related papers  (2026-01-05T14:16:11Z)
• Learning to Route LLMs from Bandit Feedback: One Policy, Many Trade-offs [69.2486294522259]
  BaRP is a Bandit Routing-feedback with Preferences approach that trains under the same partial-feedback restriction as deployment.<n> Framed as a contextual bandit over prompt features and a user preference vector, our method simulates an online feedback setting during training and adapts its routing decisions to each new prompt.
  arXiv  Detail & Related papers  (2025-10-08T18:24:59Z)
• How Robust Are Router-LLMs? Analysis of the Fragility of LLM Routing Capabilities [62.474732677086855]
  Large language model (LLM) routing has emerged as a crucial strategy for balancing computational costs with performance.<n>We propose the DSC benchmark: Diverse, Simple, and Categorized, an evaluation framework that categorizes router performance across a broad spectrum of query types.
  arXiv  Detail & Related papers  (2025-03-20T19:52:30Z)
• Routers in Vision Mixture of Experts: An Empirical Study [26.51711534240885]
  Mixture-of-Experts (MoE) models are a promising way to scale up model capacity without significantly increasing computational cost. Key component of MoEs is the router, which decides which subset of parameters (experts) process which feature embeddings (tokens)
  arXiv  Detail & Related papers  (2024-01-29T08:58:07Z)
• Patch-level Routing in Mixture-of-Experts is Provably Sample-efficient for Convolutional Neural Networks [74.68583356645276]
  In deep learning, mixture-of-experts (MoE) activates one or few experts (sub-networks) on a per-sample or per-token basis. We show for the first time that pMoE provably reduces the required number of training samples to achieve desirable generalization.
  arXiv  Detail & Related papers  (2023-06-07T00:16:10Z)
• StableMoE: Stable Routing Strategy for Mixture of Experts [109.0602120199226]
  Mixture-of-Experts (MoE) technique can scale up the model size of Transformers with an affordable computational overhead. We propose StableMoE with two training stages to address the routing fluctuation problem. Results show that StableMoE outperforms existing MoE methods in terms of both convergence speed and performance.
  arXiv  Detail & Related papers  (2022-04-18T16:48:19Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.