VLDBench
收藏资源简介:
VLDBench是由Vector Institute of AI等机构创建的,针对检测网络新闻文章中的虚假信息的一种全面的多模态基准数据集。该数据集包含来自58个新闻源的31,339对新闻文章和视觉样本,涵盖13个不同的类别。数据集通过一个严格的人工审核流程进行筛选和验证,确保了数据的高质量。VLDBench旨在为检测网络多模态内容中的虚假信息提供基准,支持单模态(仅文本)和跨模态(文本和图像)内容的评估。
VLDBench is a comprehensive multimodal benchmark dataset developed by the Vector Institute of AI and other institutions for detecting disinformation in online news articles. It contains 31,339 pairs of news articles and visual samples sourced from 58 news outlets, spanning 13 distinct categories. The dataset has been screened and validated via a rigorous manual review process to guarantee its high data quality. VLDBench aims to provide a standardized benchmark for disinformation detection in online multimodal content, supporting evaluations of both unimodal (text-only) and cross-modal (text and image) content.




