Haoran Wang 0012 Haobo Xu Ying Wang 0001 Yinhe Han 0001 CTA: Hardware-Software Co-design for Compressed Token Attention Mechanism. 429-441 2023 HPCA https://doi.org/10.1109/HPCA56546.2023.10070997 conf/hpca/2023 db/conf/hpca/hpca2023.html#WangXWH23