CHESS: Optimizing LLM Inference via Channel-Wise Thresholding and Selective Sparsification.

Junhui He, Shangyu Wu, Weidong Wen, Chun Jason Xue, Qingan Li

Browse the full EMNLP paper archive.