@inproceedings{yang-etal-2024-pyramidinfer,
    title = "{P}yramid{I}nfer: Pyramid {KV} Cache Compression for High-throughput {LLM} Inference",
    author = "Yang, Dongjie  and
      Han, Xiaodong  and
      Gao, Yan  and
      Hu, Yao  and
      Zhang, Shilin  and
      Zhao, Hai",
    editor = "Ku, Lun-Wei  and
      Martins, Andre  and
      Srikumar, Vivek",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2024",
    month = aug,
    year = "2024",
    address = "Bangkok, Thailand",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.findings-acl.195/",
    doi = "10.18653/v1/2024.findings-acl.195",
    pages = "3258--3270"
}