@inproceedings{li-etal-2025-quickllama,
    title = "{Q}uick{LL}a{MA}: Query-aware Inference Acceleration for Large Language Models",
    author = "Li, Jingyao  and
      Shi, Han  and
      Wu, Sitong  and
      Zheng, Chuanyang  and
      Li, Zhenguo  and
      Jiang, Xin  and
      Xu, Hong  and
      Jia, Jiaya",
    editor = "Rambow, Owen  and
      Wanner, Leo  and
      Apidianaki, Marianna  and
      Al-Khalifa, Hend  and
      Eugenio, Barbara Di  and
      Schockaert, Steven",
    booktitle = "Proceedings of the 31st International Conference on Computational Linguistics",
    month = jan,
    year = "2025",
    address = "Abu Dhabi, UAE",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.coling-main.34/",
    pages = "508--528"
}