English

Polysemantic Experts, Monosemantic Paths: Routing as Control in MoEs

Artificial Intelligence 2026-04-21 v1 Computation and Language Machine Learning

Abstract

An LLM's residual stream is both state and instruction: it encodes the current context and determines the next transformation. We introduce a parameter-free decomposition for Mixture-of-Experts models that splits each layer's hidden state into a control signal that causally drives routing and an orthogonal content channel invisible to the router. Across six MoE architectures, we find that models preserve surface-level features (language, token identity, position) in the content channel, while the control signal encodes an abstract function that rotates from layer to layer. Because each routing decision is low-bandwidth, this hand-off forces compositional specialization across layers. While individual experts remain polysemantic, expert paths become monosemantic, clustering tokens by semantic function across languages and surface forms. The same token (e.g., ":") follows distinct trajectories depending on whether it serves as a type annotation, an introductory colon, or a time separator. Our decomposition identifies the source of this structure: clusters in the control subspace are substantially more monosemantic than those in the full representation. As a result, the natural unit of interpretability in MoEs is not the expert but the trajectory.

Keywords

Cite

@article{arxiv.2604.17837,
  title  = {Polysemantic Experts, Monosemantic Paths: Routing as Control in MoEs},
  author = {Charles Ye and Bo Yuan and Lee Sharkey},
  journal= {arXiv preprint arXiv:2604.17837},
  year   = {2026}
}
R2 v1 2026-07-01T12:17:40.179Z