From the 1 of 2 linked papers with an AI index.
2 papers
cs.LG2026
Beyond Routing: Decoupling Expert Dispatch and Aggregation in Sparse Mixture-of-Experts
Zongfei Li
Sparse Mixture-of-Experts (MoE) routers commonly use the same scores both to select experts and to weight their already-computed outputs. We study whether these two roles, dispatch…
cs.LG2026
Dynamic Parameterization Is Not Dynamic Inference
Zongfei Li, Yuan-yih Shang, Guozhong Luo
The paper investigates whether input-dependent controller coefficients in models truly lead to dynamic inference and computational savings, introducing a Frozen-Controller Auditing…