Generalized Hebbian Algorithm

STAR: Rethinking MoE Routing as Structure-Aware Subspace Learning

Announce Type: new Abstract: Mixture-of-Experts (MoE) scales model capacity efficiently by selectively routing inputs to a specialized subset of experts. However, input-expert specialization, the core motivation of MoE, critically depends on whether the router is actually aware of input structure. In practice, MoE routing is typically implemented as a shallow linear projection with limited awareness of input representation, which often leads to unstable routing.