@inproceedings{laskar-etal-2024-query,
    title = "Query-{OPT}: Optimizing Inference of Large Language Models via Multi-Query Instructions in Meeting Summarization",
    author = "Laskar, Md Tahmid Rahman  and
      Khasanova, Elena  and
      Fu, Xue-Yong  and
      Chen, Cheng  and
      Tn, Shashi Bhushan",
    editor = "Dernoncourt, Franck  and
      Preo{\c{t}}iuc-Pietro, Daniel  and
      Shimorina, Anastasia",
    booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track",
    month = nov,
    year = "2024",
    address = "Miami, Florida, US",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.emnlp-industry.86/",
    doi = "10.18653/v1/2024.emnlp-industry.86",
    pages = "1140--1151"
}