What is Paper Banana
PaperBanana is an AI-powered tool designed to automate the creation of publication-ready academic illustrations. It transforms academic text into methodological diagrams and precise statistical plots instantly. The tool utilizes a multi-agent framework to generate scientific figures that are faithful, accurate, and aesthetically pleasing, allowing researchers to focus on their scientific work.
How to use Paper Banana
To use Paper Banana, you typically need to provide:
- Visual Intent: A brief description of the desired illustration.
- Source Context: Relevant sections from your research paper.
- Figure Caption: The caption that will appear below the generated figure.
For statistical plots, you also need to provide raw data in JSON or CSV format. The tool then generates the illustration based on these inputs.
Features of Paper Banana
- AI-powered illustration generation: Converts academic text into publication-ready diagrams and plots.
- Methodological Diagrams: Creates architectures of models, algorithmic flows, encoder-decoder frameworks, and system pipeline illustrations.
- Statistical Plots: Generates accurate statistical charts (bar charts, line graphs, scatter plots) using executable Python Matplotlib code, ensuring numerical integrity.
- Aesthetic Refinement: Enhances figures with professional quality, improving color palettes, typography, and spacing to meet publication standards.
- Multi-Agent Framework: Employs a collaborative pipeline of specialized agents (Retriever, Planner, Stylist, Visualizer, Critic) for high-fidelity results.
- Educational Infographics: Produces clear and scientifically accurate infographics for educational materials.
- Sketch Refinement: Improves rough hand-drawn sketches into professional figures.
- Real-time Web Information: Utilizes Google Search for up-to-date information to enhance image generation accuracy.
Use Cases of Paper Banana
- Generating methodological diagrams for research papers.
- Creating accurate statistical plots from raw data.
- Enhancing hand-drawn sketches into publication-quality figures.
- Producing educational infographics for supplementary materials or presentations.
- Automating the creation of figures for top-tier conferences (e.g., NeurIPS, ICML, ICLR).
FAQ
What is Paper Banana? Paper Banana is an agentic framework for the automated generation of publication-ready academic illustrations, including methodological diagrams, statistical charts, and educational infographics. It uses a multi-agent system to bridge the gap between raw scientific text and professional visual communication.
How does Paper Banana's multi-agent framework work? It coordinates five specialized agents: Retriever (locates reference visuals), Planner (translates text to visual layouts), Stylist (synthesizes aesthetic guidelines), Visualizer (renders illustrations using Nano-Banana-Pro or Python code), and Critic (performs iterative self-correction). This closed-loop architecture ensures accuracy and reliability.
How does Paper Banana ensure data accuracy in statistical plots? Instead of pixel-based rendering, Paper Banana generates executable Python Matplotlib code from raw data. This ensures that bars, data points, and scales are mathematically precise, avoiding the numerical hallucinations common in general image generators.
What is Nano-Banana-Pro? Nano-Banana-Pro is a state-of-the-art image generation model used by Paper Banana's Visualizer agent. It specializes in synthesizing scientific shapes, connectors, and icons for methodological diagrams, adhering to detailed layout instructions and maintaining visual consistency.
How is Paper Banana evaluated? Paper Banana was evaluated using Paper BananaBench, a benchmark of 292 test cases from NeurIPS 2025 publications. It reportedly outperforms leading systems like GPT-Image and Paper2Any in fidelity, conciseness, readability, and aesthetics.
Is Paper Banana open source? Yes, Paper Banana is open source under the Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) license. The codebase, model weights, training data, and benchmark are publicly available on GitHub.
What inputs does Paper Banana require to generate a diagram? It requires a Visual Intent (brief description), Source Context (relevant paper sections), and a Figure Caption. For statistical plots, raw data (JSON or CSV) is also needed.
Can Paper Banana enhance my hand-drawn sketches? Yes, Paper Banana offers an aesthetic enhancement feature that applies self-summarized aesthetic guidelines to refine color palettes, typography, and overall professional quality of rough drafts.
What types of academic figures can Paper Banana generate? It supports methodological diagrams, statistical plots, educational infographics, poster/slide visuals, and the refinement of hand-drawn sketches.
