CMMU

CMMU是一个针对中文多模态多类型问题理解和推理的新型基准，由北京人工智能研究院创建。该数据集包含3603个问题，覆盖从小学到高中的7个学科知识。问题类型多样，包括选择题、多选题和填空题，旨在全面评估多模态大型语言模型（MLLMs）的理解和推理能力。CMMU不仅提供答案，还提供答案解释，以帮助理解和评估模型的推理过程。数据集的应用领域广泛，特别是在教育和人工智能领域，用于评估和提升模型的专业知识理解和复杂问题解决能力。

CMMU is a novel benchmark for Chinese multimodal and multi-type question understanding and reasoning, developed by the Beijing Academy of Artificial Intelligence. This dataset includes 3603 questions covering 7 academic subjects ranging from primary to high school levels. The question types are diverse, including single-choice questions, multiple-choice questions and fill-in-the-blank questions, aiming to comprehensively evaluate the understanding and reasoning capabilities of multimodal large language models (MLLMs). CMMU not only provides answers but also corresponding answer explanations to facilitate understanding and assessment of the model's reasoning process. The dataset has a wide range of application scenarios, particularly in the fields of education and artificial intelligence, for evaluating and enhancing the professional knowledge comprehension and complex problem-solving abilities of models.

搜集汇总

背景与挑战

背景概述

CMMU是一个由北京人工智能研究院创建的中文多模态多类型问题理解和推理基准，包含3603个问题，覆盖从小学到高中的7个学科，问题类型包括选择题、多选题和填空题，旨在评估多模态大型语言模型的理解和推理能力。数据集不仅提供答案，还提供答案解释，以帮助分析模型的推理过程，主要应用于教育和人工智能领域，用于提升模型的专业知识理解和复杂问题解决能力。