一作ACL 2026Findings· Findings of ACL
TENP : Trapezoidal Expert Neuron Pruning For Mixture-of-Experts
Jiangyang He · Shaolin Zhu · Deyi Xiong
摘要
Mixture-of-Experts large language models (LLMs) scale efficiently through sparse activation, yet their deployment is fundamentally constrained by the large static parameter footprint of experts. Existing compression approaches either remove entire experts, disrupting routing topology and harming...