AutoFigure: Generating and Refining Publication-Ready Scientific Illustrations

2026-02-03

AI Summary

This work addresses the time-consuming and labor-intensive process of manually creating high-quality scientific illustrations, a common bottleneck in both academic research and industry. We propose AutoFigure, the first agent-based framework capable of automatically generating and optimizing publication-ready scientific figures from long-form scientific text. Employing an end-to-end architecture, AutoFigure is the first to produce structurally coherent and aesthetically refined illustrations directly from extended textual inputs, enhanced by integrated mechanisms for reasoning, reorganization, and validation to ensure output quality. To support evaluation and future research, we introduce FigureBench, a benchmark dataset comprising 3,300 figure–text pairs. Experimental results demonstrate that AutoFigure significantly outperforms existing baselines, consistently generating figures that meet publication standards. The code, dataset, and demonstration platform are publicly released.

Technology Category

• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Natural Language Processing: Generation
• Multiagent Systems: Agent-Based Simulation and Emergent Behavior

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

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.

Problem

• scientific illustrations
• manual creation bottleneck
• publication-ready figures
• scientific communication

Innovation

• scientific illustration generation
• agentic framework
• FigureBench
• publication-ready visualization
• text-to-figure synthesis