作者位次 #8/10ACL 2026Findings· Findings of ACL
Exploring Layer-wise Information Effectiveness for Post-Training Quantization in Small Language Models
He Xiao · Qingyao Yang · Dirui Xie · Wendong XU · Zunhai Su · Runming Yang · Haobo Liu · Wenyong Zhou · Zhengwu Liu · Ngai Wong
摘要
Large language models with billions of parameters are often over-provisioned: many layers contribute little unique information yet dominate the memory and energy footprint during inference. We present LieQ (Layer-wise information effectiveness Quantization), a hardware-native, metric-driven post-...