Multi-Physics
收藏资源简介:
Multi-Physics是一个针对中国物理问题的多模态语言模型(MLLMs)推理的综合基准数据集,由香港中文大学(深圳)科学与工程学院的研究团队构建。该数据集包含了1,412个与图像关联的多选题,涵盖11个高中物理学科,并分为5个难度等级。数据集通过双评价框架评估了20种不同的MLLMs,分析了最终答案的准确性和思考过程的完整性。此外,数据集还系统地研究了难度等级和视觉信息对MLLMs推理的影响。
Multi-Physics is a comprehensive benchmark dataset for multimodal large language model (MLLM) reasoning targeting Chinese physical problems, developed by the research team from the School of Science and Engineering at The Chinese University of Hong Kong, Shenzhen. The dataset comprises 1,412 image-associated multiple-choice questions spanning 11 high school physics disciplines, and is divided into 5 difficulty levels. It uses a dual evaluation framework to assess 20 distinct MLLMs, analyzing the accuracy of the models' final answers and the completeness of their reasoning processes. Additionally, the dataset systematically explores the impacts of difficulty levels and visual information on MLLM reasoning.

- 1通过香港中文大学(深圳)科学与工程学院 · 2025年



