SCAFFOLD: A Large-Scale Structured Dataset of Computer Science Research Figures with Diagram QA and Chain-of-Thought Reasoning Traces

SCAFFOLD：一个包含图表问答与思维链推理轨迹的大规模计算机科学研究图表结构化数据集

Computer science papers rely heavily on diagrams: architecture drawings, system flowcharts, and pipeline schematics that often carry more information than the text around them. 计算机科学论文高度依赖图表：架构图、系统流程图和管线示意图，这些图表往往比周围的文本承载了更多的信息。

There is currently no public dataset that pairs this specific kind of figure with captions, context, questions, answers, and step-by-step reasoning, which is exactly what is needed to train a vision-language model to understand them. 目前尚无公开数据集将此类特定图表与标题、上下文、问题、答案及逐步推理过程进行配对，而这正是训练视觉语言模型以理解这些图表所必需的。

We present SCAFFOLD, a large-scale structured dataset of computer science research figures with diagram QA and Chain-of-Thought reasoning traces. 我们提出了 SCAFFOLD，这是一个包含图表问答（QA）和思维链（Chain-of-Thought）推理轨迹的大规模计算机科学研究图表结构化数据集。

This dataset consists of (image, caption, context, question-answer, chain-of-thought) tuples from arXiv computer science papers prepared using layout detection and PDF parsing, with an AI-assisted question-generation step. 该数据集由来自 arXiv 计算机科学论文的（图像、标题、上下文、问答、思维链）元组组成，通过布局检测和 PDF 解析技术构建，并辅以 AI 辅助的问题生成步骤。

The resulting large-sized SCAFFOLD-157K dataset spans 3,058 papers with 29,887 figures (157,387 pairs), a medium-sized SCAFFOLD-37K dataset (36,797 pairs), and a small-sized SCAFFOLD-12K dataset (12,000 pairs). 最终生成的大规模 SCAFFOLD-157K 数据集涵盖了 3,058 篇论文中的 29,887 张图表（共 157,387 个配对），此外还包括中等规模的 SCAFFOLD-37K 数据集（36,797 个配对）和小规模的 SCAFFOLD-12K 数据集（12,000 个配对）。

We used SCAFFOLD-12K for baseline experiments on Qwen2.5-VL-3B-Instruct. 我们使用 SCAFFOLD-12K 对 Qwen2.5-VL-3B-Instruct 模型进行了基准实验。