MMMU
收藏资源简介:
MMMU是一个大规模的多学科多模态理解和推理基准,用于评估多模态模型在需要大学水平学科知识和深思熟虑推理的大量多学科任务上的表现。该数据集包含从大学考试、测验和教科书中精心收集的11.5K个多模态问题,涵盖艺术与设计、商业、科学、健康与医学、人文与社会科学以及技术和工程六个核心学科。这些问题跨越30个学科和183个子领域,包含30种高度异质的图像类型,如图表、图解、地图、表格、音乐表和化学结构。
MMMU is a large-scale multidisciplinary multimodal understanding and reasoning benchmark for evaluating the performance of multimodal models on a wide range of multidisciplinary tasks that require college-level disciplinary knowledge and deliberate reasoning. This dataset contains 11.5K multimodal questions carefully collected from college-level examinations, quizzes and textbooks, covering six core disciplines: art and design, business, science, health and medicine, humanities and social sciences, as well as technology and engineering. These questions span 30 disciplines and 183 sub-fields, and include 30 highly heterogeneous image types such as charts, diagrams, maps, tables, musical scores and chemical structures.

- 1MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI · 2023年



