MASSW

MASSW是由密歇根大学开发的一个大规模、结构化的科学工作流程数据集，包含超过152,000篇来自17个顶级计算机科学会议的同行评审出版物。该数据集通过大型语言模型自动提取了科学工作流程的五个核心方面：背景、关键思想、方法、结果和预期影响，以支持多种机器学习任务，如思想生成和结果预测。MASSW旨在通过提供一个结构化、准确的数据资源，帮助研究人员开发和评估优化科学工作流程的创新AI方法，从而促进科学领域的更快创新。

MASSW is a large-scale, structured scientific workflow dataset developed by the University of Michigan. It contains over 152,000 peer-reviewed publications from 17 top-tier computer science conferences. Five core aspects of scientific workflows, namely background, key ideas, methods, results and expected impacts, are automatically extracted via large language models from this dataset to support a variety of machine learning tasks such as idea generation and result prediction. MASSW aims to help researchers develop and evaluate innovative AI methods for optimizing scientific workflows by providing a structured and accurate data resource, thereby promoting faster innovation in scientific fields.