arXiv:2608.29681 · 多模态
MMMMM: A Unified Taxonomy for Investigating the Mechanisms of Multilingual MultiModal Misinformation
MMMMM：用于研究多语言多模态虚假信息机制的统一分类法
- 类型：arxiv
- 标识：2608.29681
- 链接：https://arxiv.org/abs/2608.29681
- 主分类：multimodal
- 形态：method
- TLDR：Multimodal misinformation on social media is highly prevalent, potent, and harmful, yet difficult to detect and counter, and still poorly understood compared to its text-only counterpart. Research on the properties and deceptive strategies of multimodal misinformation is hindered by a lack of taxonomies grounded in real-world contexts and by the limitations of current multimodal machine learning models, which prevent the automation of annotation and analysis at scale. We address these shortcomings in three steps. First, we collect a large-scale, high-quality dataset of real-world misinformatio
- 待LLM分类：否
- 标题中文：MMMMM：用于研究多语言多模态虚假信息机制的统一分类法
- TLDR中文：社交媒体上的多模态虚假信息极为普遍、影响强且危害大，但相比纯文本虚假信息更难以检测和应对，也更难被理解。对多模态虚假信息的特征与欺骗策略的研究，受限于缺乏基于真实场景的分类体系，以及当前多模态机器学习模型无法支撑大规模自动化标注与分析。我们分三步解决这些问题。首先，我们构建了一个大规模、高质量的真实世界虚假信息数据集
- 来源文件：
- /inbox/tom/_candidates/2026-09-02-agent-rag-longcontext-candidates.json