MBZUAI/EXAMS-V

EXAMS-V是一个多语言、多模态数据集，旨在评估和基准AI系统，尤其是视觉语言模型(VLMs)的视觉推理能力。该数据集包含了24,856个从真实学校考试和其他教育来源收集的多项选择题(MCQs)，所有问题都以图像形式呈现。这些问题不仅包含文本，还包括表格、图表和数学内容，这使得EXAMS-V成为测试模型处理视觉和结构化信息能力的强大基准。问题以13种不同的语言编写，涵盖多个学科领域。数据集从不同国家和教育体系的真实学校考试中策划而来，具有独特的地区特定知识、多样的问题格式和多种语言内容。回答EXAMS-V中的问题不仅仅是阅读，模型还需要理解视觉布局、解释图表和符号，并对文本和视觉进行推理。

EXAMS-V is a multilingual, multimodal dataset created to evaluate and benchmark the visual reasoning abilities of AI systems, especially Vision-Language Models (VLMs). The dataset consists of 24,856 multiple-choice questions (MCQs) collected from real school exams and other educational sources, all presented as images. These images include not only text but also tables, graphs, and mathematical content, making EXAMS-V a strong benchmark for testing how well models can handle visual and structured information. The questions are written in 13 different languages and cover multiple subject domains. The dataset is curated from real school exams across different countries and educational systems, offering a unique mix of region-specific knowledge, varied question formats, and multilingual content. Answering questions in EXAMS-V requires models to understand visual layout, interpret diagrams and symbols, and reason over both text and visuals.