作者位次 #4/7ACL 2026Poster· Natural Language Processing
Profiling-Free Mixed-Precision Quantization for M o E LLM s via Fuzzy Rule Interpolation
Huachen Qi · Ruiyu Zhuo · Bowen Shi · Xiang Chang · Fei Chao · Changjing Shang · Qiang Shen
摘要
Large Language Models continue to scale in size and capability, driving substantial computational and memory demands.Mixture-of-Experts (MoE) architectures alleviate this cost by activating only a sparse subset of experts per token, enabling efficient scaling without proportional increases in inf...