作者位次 #2/9ACL 2026Findings· Findings of ACL
From Outcome to Process: Optimizing M o E Load Balancing with MCTS
Wenjun Ke · Hengyuan Xu · Ziyu Shang · Yao He · Jiahao Wang · Zijie Xu · Peng Wang · Yuhang Lou · Jiajun Liu
摘要
Mixture of Experts (MoE) dynamically routes inputs to specialized expert networks, enabling large language models to scale capacity with low inference overhead. To further improve MoE’s parameter efficiency in resource-constrained scenarios, LoRA–MoE integrates LoRA for lightweight adaptation whi...