arXiv · Statistics Machine Learning· Yanlong Zhao, Xiaoyuan Cheng, Huihang Liu, Baihua He, Xinyu Zhang, Harrison Bo Hua Zhu, Wenlong Chen, Li Zeng, Zhuo Sun·· 4 hr agoAI score33
低重构损失为何反而有害:面向低位 LLM 量化的分布鲁棒精炼
When Lower Reconstruction Loss Hurts: Distributionally Robust Refinement for Low-Bit LLM Quantization
AI brief
论文提出 Distributionally Robust Quantization(DRQ),一种针对 weight-only PTQ 的后处理精炼框架,在受限输入激活分布集合上最小化最坏情况重构损失。
Source: arXiv · Statistics Machine Learning · arxiv.org