AutoFigure：生成与优化可发表级科学插图

AutoFigure: Generating and Refining Publication-Ready Scientific Illustrations

February 3, 2026

作者: Minjun Zhu, Zhen Lin, Yixuan Weng, Panzhong Lu, Qiujie Xie, Yifan Wei, Sifan Liu, Qiyao Sun, Yue Zhang

cs.AI

摘要

高质量科学插图对于有效传达复杂科技概念至关重要，然而其人工创作始终是学术界与工业界公认的瓶颈。我们推出FigureBench——首个基于长篇科学文本生成插图的大规模基准数据集，包含3,300组高质量科学文本-插图配对样本，涵盖科研论文、综述、博客及教材中的多样化文转图任务。此外，我们提出AutoFigure——首个基于长篇科学文本自动生成高质量插画的智能体框架。该框架在最终渲染前会进行深度思考、要素重组与多轮验证，生成兼具结构合理性与美学精炼度的布局方案，输出结构完整且视觉精美的科学插图。依托FigureBench提供的高质量数据，我们开展大量实验对比AutoFigure与多种基线方法的性能。结果表明AutoFigure持续超越所有基线方法，能生成达到出版标准的科学插图。代码、数据集及HuggingFace空间已发布于https://github.com/ResearAI/AutoFigure。

English

High-quality scientific illustrations are crucial for effectively communicating complex scientific and technical concepts, yet their manual creation remains a well-recognized bottleneck in both academia and industry. We present FigureBench, the first large-scale benchmark for generating scientific illustrations from long-form scientific texts. It contains 3,300 high-quality scientific text-figure pairs, covering diverse text-to-illustration tasks from scientific papers, surveys, blogs, and textbooks. Moreover, we propose AutoFigure, the first agentic framework that automatically generates high-quality scientific illustrations based on long-form scientific text. Specifically, before rendering the final result, AutoFigure engages in extensive thinking, recombination, and validation to produce a layout that is both structurally sound and aesthetically refined, outputting a scientific illustration that achieves both structural completeness and aesthetic appeal. Leveraging the high-quality data from FigureBench, we conduct extensive experiments to test the performance of AutoFigure against various baseline methods. The results demonstrate that AutoFigure consistently surpasses all baseline methods, producing publication-ready scientific illustrations. The code, dataset and huggingface space are released in https://github.com/ResearAI/AutoFigure.