Increasing active parameters per token in MOE (Qwen 35B A4B+) reduce reasoning token by 8.5% - and you don't need to train or finetune!

I want to share a short paper just published exploring a simple but surprisingly effective optimization for sparse MoE reasoning models. The idea: Instead of retraining anything, we just tweak the router at runtime. Specifically, we expand the expert selection budget (N≥KN≥K) only in the late transf