中国移动学科视觉问答数据集

本数据集为多学科领域视觉问答（VQA）专属图文配对数据集，完整覆盖艺术、商科、理科、医学、人文社科、工科六大学科领域。每条样本包含学科题目或场景图像、关联问答问题、标准精准答案、详细知识点解析及学科分类标签，问题内嵌图像定位标记，解析完整覆盖考点与推理逻辑，聚焦几何图形、电路原理、化学实验、函数图像等典型学科内容，实现图像信息、问答交互与学科知识的深度匹配，数据格式统一、内容严谨规范。This is a dedicated image-text paired dataset for multi-disciplinary Visual Question Answering (VQA), which comprehensively covers six academic disciplines: arts, business, natural sciences, medicine, humanities and social sciences, and engineering. Each sample contains subject-specific questions or scenario images, associated question-answer pairs, standard and precise answers, detailed knowledge point explanations, and discipline classification tags. The questions are embedded with image localization markers, and the explanations fully cover examination points and reasoning logic, focusing on typical disciplinary contents such as geometric figures, circuit principles, chemical experiments, and function graphs. This dataset achieves deep matching among image information, question-answer interaction and disciplinary knowledge, with unified data format and rigorous and standardized content.

搜集汇总

背景与挑战

背景概述

该数据集是一个专注于多学科领域的视觉问答（VQA）图文配对数据集，涵盖艺术、商科、理科、医学、人文社科和工科六大学科，包含约55万条数据，每条数据均提供图像、问题、答案、解析和学科标签。它适用于多模态教学、智能解题及AI教育产品开发等场景，旨在提升AI在学科知识理解与推理方面的能力。