Paper2Poster – An academic poster generation framework developed by the University of Waterloo, the National University of Singapore and the University of Oxford

What is Paper2Poster

Paper2Poster is an innovative academic framework developed by institutions including the University of Waterloo (Canada) and the National University of Singapore. It leverages multimodal automation technology to generate academic posters directly from scientific papers. Paper2Poster introduces PosterAgent, a top-down multi-agent system that supports compressing lengthy research papers into structured visual posters. The system consists of three main components: a Parser, a Planner, and a Painter–Commenter loop, enabling efficient poster generation. Additionally, Paper2Poster introduces an evaluation method called PaperQuiz, which simulates reader questions to assess the poster’s ability to convey core content. Paper2Poster demonstrates excellent performance in visual quality and textual coherence, significantly improving generation efficiency and providing a cost-effective solution for academic poster creation.

Main Features of Paper2Poster

• Long-Text Compression: Condenses multi-page scientific papers into single-page posters while preserving the core content.

• Multimodal Content Handling: Extracts text, figures, and images from papers and integrates them effectively into the poster.

• Layout Optimization: Generates aesthetically pleasing and logically organized layouts to ensure optimal use of limited poster space.

• Visual Quality Enhancement: Improves the poster’s visual appeal and readability through a visual feedback mechanism.

• Evaluation and Optimization: Uses PaperQuiz to assess how well the poster communicates key ideas, and refines the output based on feedback.

Technical Architecture of Paper2Poster

• Parser: Uses tools such as MARKER and DOCLING to convert PDFs into Markdown format, then utilizes LLMs to generate a structured asset library in JSON format.

• Planner: Aligns text and visual elements from the asset library and generates a binary tree layout. Based on this layout strategy, it estimates panel sizes according to content length, ensuring reading order and spatial balance. It also uses LLMs for semantic matching, aligning visual elements with the most relevant textual sections.

• Painter–Commenter: Responsible for generating content in each panel and optimizing the layout using a visual feedback mechanism. The Painter aligns text and images and generates executable code rendered via the python-pptx library. The Commenter, a vision-language model (VLM), provides layout feedback using zoom-in reference prompts to ensure content fits well and is logically arranged.

Project Links for Paper2Poster

• Official Website: https://paper2poster.github.io/

• GitHub Repository: https://github.com/Paper2Poster/Paper2Poster

• HuggingFace Dataset: https://huggingface.co/datasets/Paper2Poster/Paper2Poster

• arXiv Technical Paper: https://arxiv.org/pdf/2505.21497

Application Scenarios of Paper2Poster

• Academic Conferences: Helps researchers quickly convert their papers into posters for conference presentations, saving time and effort.

• Academic Presentations: Posters can serve as supplementary materials for academic talks, aiding audience comprehension.

• Research Showcase: Used in research institutions or laboratories to display recent research outcomes, facilitating peer exchange and learning.

• Education: Enables educators to create teaching posters that help students visually understand complex academic concepts.

• Online Academic Platforms: Provides automated poster generation tools for academic platforms, enhancing user experience and promoting academic communication.