作者位次 #2/8ACL 2026Poster· Natural Language Processing
Theory-optimal Quantization Based on Flatness
Xiusheng Huang · Zhe Li · Xuanwu Yin · Lu Wang · Yequan Wang · Dong Li · Emad Barsoum · Kang Liu
摘要
Post-training quantization has emerged as a widely adopted technique for compressing and accelerating the inference of Large Language Models (LLMs). The primary challenges in LLMs quantization stem from activation outliers, which significantly degrade model performance especially at lower bit pre...