SciPG: A New Benchmark and Approach for Layout-aware Scientific Poster Generation
Keywords: Scientific poster generation, multimodal extraction, multimodal generation
TL;DR: A New Benchmark and Approach for Scientific Poster Generation
Abstract: Scientific posters are an effective and expressive medium for conveying the core ideas of academic papers, facilitating the communication of research techniques. However, creating high-quality scientific posters is a complex and time-consuming task that requires advanced skills to summarize key concepts and arrange them logically and visually appealingly. Previous studies have primarily focused on either content extraction or the layout and composition of posters, often relying on small-scale datasets. The scarcity of large, publicly available datasets has further limited advancements in this field.
In this paper, we introduce a new task called layout-aware scientific poster generation (LayoutSciPG), which aims to generate flexible posters from scientific papers through integrated automatic content extraction and layout design.
To achieve this, we first build a large-scale dataset containing over 10,000 pairs of scientific papers and their corresponding posters. We then propose a multimodal extractor-generator framework, which employs a multimodal extractor to retrieve key text and image elements from the papers and designs an interactive generator with an adaptive memory mechanism to seamlessly paraphrase the extracted content and generate a structured layout. This approach effectively tackles challenges related to GPU memory consumption and long-term dependencies when handling the lengthy inputs (scientific papers) and outputs (posters). Finally, both qualitative and quantitative evaluations demonstrate the effectiveness of our approach while highlighting remaining challenges.
Primary Area: datasets and benchmarks
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 8678
Loading