Fetching the paper…

A Simple and Effective $L_2$ Norm-Based Strategy for KV Cache Compression · Around