@inproceedings{huang-etal-2026-mobilellm,
    title = "{M}obile{LLM}-Flash: Latency-Guided On-Device {LLM} Design for Industry Scale Deployment",
    author = "Huang, Hanxian  and
      Fedorov, Igor  and
      Gromov, Andrey  and
      Beckerman, Bernard  and
      Suda, Naveen  and
      Eriksson, David  and
      Balandat, Maximilian  and
      Conway, Rylan  and
      Huber, Patrick  and
      Sankar, Chinnadhurai  and
      Dalmia, Ayushi  and
      Liu, Zechun  and
      Wu, Lemeng  and
      Elgamal, Tarek  and
      Sagar, Adithya  and
      Chandra, Vikas  and
      Krishnamoorthi, Raghuraman",
    editor = "Li, Yunyao  and
      Rehm, Georg  and
      Tu, Mei",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics ({ACL} 2026)",
    month = jul,
    year = "2026",
    address = "San Diego, California, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-industry.51/",
    pages = "749--760",
    ISBN = "979-8-89176-394-4"
}