DesignQA
收藏资源简介:
DesignQA是一个专为评估大型语言模型在理解工程技术文档方面的能力而设计的多模态基准数据集。该数据集由麻省理工学院与Autodesk研究团队合作开发,包含1451个问题,这些问题基于Formula SAE学生竞赛的文本设计要求、CAD图像和工程图纸等多模态数据。DesignQA旨在通过模拟真实世界工程挑战,测试模型在处理复杂工程设计要求时的理解和应用能力。数据集的应用领域主要集中在工程设计自动化,旨在通过AI辅助,使人类设计师能更快、更有效地创造出更优质的产品。
DesignQA is a multimodal benchmark dataset specifically developed to evaluate the capabilities of large language models (LLMs) in understanding engineering technical documents. It was collaboratively created by the Massachusetts Institute of Technology (MIT) and the Autodesk Research team, and contains 1,451 questions based on multimodal data including text design requirements, CAD images, and engineering drawings from the Formula SAE student competition. DesignQA aims to test models' comprehension and application abilities when handling complex engineering design requirements by simulating real-world engineering challenges. Its primary application domain is engineering design automation, with the objective of enabling human designers to produce higher-quality products faster and more efficiently through AI assistance.

- 1DesignQA: A Multimodal Benchmark for Evaluating Large Language Models' Understanding of Engineering Documentation麻省理工学院 · 2024年



