IWR-Bench/IWR-Bench

IWR-Bench是一个用于评估大型视觉语言模型在重构用户交互视频中的网页重建能力的基准数据集。它包含了用户与网页交互的视频以及与之相关的文本描述，支持video-text-to-text和visual-question-answering任务。数据集的特征包括交互复杂性、视觉复杂性、领域、子领域、录制细节、源网站、动作序列和视频路径等信息。

IWR-Bench is a benchmark dataset for evaluating the webpage reconstruction ability of large-scale visual language models from user interaction videos. It includes videos of user interactions with web pages and associated text descriptions, supporting video-text-to-text and visual-question-answering tasks. The dataset features include interaction complexity, visual complexity, domain, sub-domain, recording details, source website, action sequence, and video path information.