@inproceedings{wang-etal-2025-filter,
    title = "Filter-And-Refine: A {MLLM} Based Cascade System for Industrial-Scale Video Content Moderation",
    author = "Wang, Zixuan  and
      Shi, Jinghao  and
      Liang, Hanzhong  and
      Shen, Xiang  and
      Wen, Vera  and
      Chen, Zhiqian  and
      Wu, Yifan  and
      Zhang, Zhixin  and
      Xiong, Hongyu",
    editor = "Rehm, Georg  and
      Li, Yunyao",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/landing_page/2025.acl-industry.62/",
    pages = "873--880",
    ISBN = "979-8-89176-288-6",
    abstract = "Effective content moderation is essential for video platforms to safeguard user experience and uphold community standards. While traditional video classification models effectively handle well-defined moderation tasks, they struggle with complicated scenarios such as implicit harmful content and contextual ambiguity. Multimodal large language models (MLLMs) offer a promising solution to these limitations with their superior cross-modal reasoning and contextual understanding. However, two key challenges hinder their industrial adoption. First, the high computational cost of MLLMs makes full-scale deployment impractical. Second, adapting generative models for discriminative classification remains an open research problem. In this paper, we first introduce an efficient method to transform a generative MLLM into a multimodal classifier using minimal discriminative training data. To enable industry-scale deployment, we then propose a router-ranking cascade system that integrates MLLMs with a lightweight router model. Offline experiments demonstrate that our MLLM-based approach improves F1 score by 66.50{\%} over traditional classifiers while requiring only 2{\%} of the fine-tuning data. Online evaluations show that our system increases automatic content moderation volume by 41{\%}, while the cascading deployment reduces computational cost to only 1.5{\%} of direct full-scale deployment."
}