DiagramGenBenchmark
收藏资源简介:
DiagramGenBenchmark是由中国科学院大学和西湖大学共同创建的一个用于评估结构化图表生成和编辑的综合性数据集。该数据集包含八种不同类型的图表,如流程图、模型架构图和思维导图,旨在为文本到图表生成任务提供坚实的基础。数据集的内容涵盖了从简单到复杂的多种图表类型,确保了评估的全面性。创建过程中,研究团队通过严格的筛选和标注,确保了数据集的高质量和多样性。该数据集主要应用于教育、科研和工业领域,旨在解决复杂信息的高效可视化表达问题,提升图表生成和编辑的准确性和灵活性。
DiagramGenBenchmark is a comprehensive dataset jointly created by the University of Chinese Academy of Sciences and Westlake University for evaluating structured diagram generation and editing. It encompasses eight distinct types of diagrams, including flowcharts, model architecture diagrams and mind maps, aiming to provide a robust foundation for text-to-diagram generation tasks. The dataset covers a wide range of diagram types spanning from simple to complex, ensuring the comprehensiveness of the evaluation. During the dataset development process, the research team implemented strict screening and annotation protocols to guarantee its high quality and diversity. This dataset is primarily applied in education, scientific research and industrial fields, with the goal of addressing efficient visual expression of complex information and improving the accuracy and flexibility of diagram generation and editing.

- 1From Words to Structured Visuals: A Benchmark and Framework for Text-to-Diagram Generation and Editing中国科学院大学 西湖大学 · 2024年



