TIGER-Lab/VisualWebInstruct-Recall

VisualWebInstruct-Recall数据集是从Google搜索中根据种子图像回忆起来的，用于多模态指令数据的大规模扩展。数据集包含了问题、答案以及与之相关的图像，旨在支持问答和图像文本到文本的任务。数据集具有多模态特性，包括文本和图像信息，并提供了训练集划分。

The VisualWebInstruct-Recall dataset is recalled from Google Search based on seed images for large-scale expansion of multimodal instruction data. The dataset includes questions, answers, and related images, designed to support question-answering and image-text-to-text tasks. The dataset is multimodal, incorporating both text and image information, and provides a training set split.