Routing-Free Mixture-of-Experts
This addresses the problem of rigid routing in MoE models for machine learning researchers, representing a novel method rather than an incremental improvement.
The paper tackles the rigid inductive biases in standard Mixture-of-Experts models by proposing Routing-Free MoE, which eliminates centralized routing mechanisms and instead encapsulates activation functionalities within individual experts optimized through continuous gradient flow. Experiments show it consistently outperforms baselines with better scalability and robustness.
Standard Mixture-of-Experts (MoE) models rely on centralized routing mechanisms that introduce rigid inductive biases. We propose Routing-Free MoE which eliminates any hard-coded centralized designs including external routers, Softmax, Top-K and load balancing, instead encapsulating all activation functionalities within individual experts and directly optimized through continuous gradient flow, enabling each expert to determine its activation entirely on its own. We introduce a unified adaptive load-balancing framework to simultaneously optimize both expert-balancing and token-balancing objectives through a configurable interpolation, allowing flexible and customizable resource allocation. Extensive experiments show that Routing-Free MoE can consistently outperform baselines with better scalability and robustness. We analyze its behavior in detail and offer insights that may facilitate future MoE design ad optimization.