作者位次 #3/12ACL 2026Poster· Natural Language Processing
R oute M o A : Dynamic Routing without Pre-Inference Boosts Efficient Mixture-of-Agents
Jize Wang · Han Wu · Zhiyuan You · Yiming Song · Yijun Wang · Zifei Shan · Yining Li · Songyang Zhang · Xinyi Le · Cailian Chen · Xinping Guan · Dacheng Tao
摘要
Mixture-of-Agents (MoA) improves LLM performance through layered collaboration, but its dense topology raises costs and latency. Existing methods employ LLM judges to filter responses, yet still require all models to perform inference before judging, failing to cut costs effectively. They also la...