作者位次 #6/11ACL 2026Poster· Natural Language Processing
Uncertainty-Aware Routing for Principled Alignment with M o E Dynamics
Yilong Chen · Junyuan Shang · Yuchen Feng · Zhenyu Zhang · Naibin Gu · Ziqi Wang · Tingwen Liu · Shuohuan Wang · YU SUN · hua wu · Haifeng Wang
摘要
Mixture-of-Experts (MoE) is a cornerstone for scaling LLMs, yet its training dynamics remain poorly understood, often leading to sub-optimal specialization. Moving beyond static routing, we present a systematic study of the MoE lifecycle using Helmholtz Free Energyand Router Entropy. We identify...