@inproceedings{wang-etal-2025-effivlm,
    title = "{E}ffi{VLM}-{BENCH}: A Comprehensive Benchmark for Evaluating Training-Free Acceleration in Large Vision-Language Models",
    author = "Wang, Zekun  and
      Ma, MingHua  and
      Wang, Zexin  and
      Mu, Rongchuan  and
      Shan, Liping  and
      Liu, Ming  and
      Qin, Bing",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingestion-acl-25/2025.acl-long.1242/",
    pages = "25546--25572",
    ISBN = "979-8-89176-251-0",
    abstract = "Large Vision-Language Models (LVLMs) have achieved remarkable success, yet their significant computational demands hinder practicaldeployment. While efforts to improve LVLM efficiency are growing, existing methods lack comprehensive evaluation across diverse backbones, benchmarks, and metrics. In this work, we systematically evaluate mainstream acceleration techniques for LVLMs, categorized into token and parameter compression. We introduce EffiVLM-BENCH, a unified framework for assessing not only absolute performance but also generalization and loyalty, while exploring Pareto-optimal trade-offs. Our extensive experiments and in-depth analyses offer insights into optimal strategies for accelerating LVLMs. We open-source code and recipes for EffiVLM-BENCH to foster future research."
}