lmms-lab/VisualWebBench

该数据集是一个多配置的数据集，包含了动作标注、动作预测、元素标注、元素OCR、标题OCR、网络图片字幕和网页问答等不同类型的任务。每个配置都有特定的特征，如图片、问题描述、答案等，并且包含了测试集。

This dataset is a multi-configuration dataset that includes tasks such as action annotation, action prediction, element annotation, element OCR, heading OCR, web image captioning, and web QA. Each configuration has specific features like images, question descriptions, answers, etc., and includes a test set.