Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SciPostGen: Bridging the Gap between Scientific Papers and Poster Layouts

Shun Inadumi, Shohei Tanaka, Tosho Hirasawa, Atsushi Hashimoto, Koichiro Yoshino, Yoshitaka Ushiku

TL;DR

This work introduces SciPostGen, a large-scale dataset of 18,097 paper–poster pairs with rich OCR and layout annotations, enabling systematic analysis of how scientific papers map to poster layouts. It reveals correlations between paper structures (text, figures, tables) and poster layout elements, motivating a Retrieval-Augmented Poster Layout Generation framework that combines a contrastively trained layout retriever with a GPT-5-based layout generator. The approach is evaluated under automatic and semi-automatic settings, showing that retrieved layouts align with paper structures and that prompting the generator with retrieved examples and paper constraints improves layout quality, while still facing trade-offs in element counts. By providing both the dataset and a practical generation framework, the work lays groundwork for data-driven understanding and automatic creation of scientifically faithful posters.

Abstract

As the number of scientific papers continues to grow, there is a demand for approaches that can effectively convey research findings, with posters serving as a key medium for presenting paper contents. Poster layouts determine how effectively research is communicated and understood, highlighting their growing importance. In particular, a gap remains in understanding how papers correspond to the layouts that present them, which calls for datasets with paired annotations at scale. To bridge this gap, we introduce SciPostGen, a large-scale dataset for understanding and generating poster layouts from scientific papers. Our analyses based on SciPostGen show that paper structures are associated with the number of layout elements in posters. Based on this insight, we explore a framework, Retrieval-Augmented Poster Layout Generation, which retrieves layouts consistent with a given paper and uses them as guidance for layout generation. We conducted experiments under two conditions: with and without layout constraints typically specified by poster creators. The results show that the retriever estimates layouts aligned with paper structures, and our framework generates layouts that also satisfy given constraints.

SciPostGen: Bridging the Gap between Scientific Papers and Poster Layouts

TL;DR

This work introduces SciPostGen, a large-scale dataset of 18,097 paper–poster pairs with rich OCR and layout annotations, enabling systematic analysis of how scientific papers map to poster layouts. It reveals correlations between paper structures (text, figures, tables) and poster layout elements, motivating a Retrieval-Augmented Poster Layout Generation framework that combines a contrastively trained layout retriever with a GPT-5-based layout generator. The approach is evaluated under automatic and semi-automatic settings, showing that retrieved layouts align with paper structures and that prompting the generator with retrieved examples and paper constraints improves layout quality, while still facing trade-offs in element counts. By providing both the dataset and a practical generation framework, the work lays groundwork for data-driven understanding and automatic creation of scientifically faithful posters.

Abstract

As the number of scientific papers continues to grow, there is a demand for approaches that can effectively convey research findings, with posters serving as a key medium for presenting paper contents. Poster layouts determine how effectively research is communicated and understood, highlighting their growing importance. In particular, a gap remains in understanding how papers correspond to the layouts that present them, which calls for datasets with paired annotations at scale. To bridge this gap, we introduce SciPostGen, a large-scale dataset for understanding and generating poster layouts from scientific papers. Our analyses based on SciPostGen show that paper structures are associated with the number of layout elements in posters. Based on this insight, we explore a framework, Retrieval-Augmented Poster Layout Generation, which retrieves layouts consistent with a given paper and uses them as guidance for layout generation. We conducted experiments under two conditions: with and without layout constraints typically specified by poster creators. The results show that the retriever estimates layouts aligned with paper structures, and our framework generates layouts that also satisfy given constraints.

Paper Structure

This paper contains 48 sections, 2 equations, 18 figures, 12 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Scientific Poster Generation.
  4. Layout Generation.
  5. SciPostGen Dataset
  6. Annotation Overview
  7. Data Construction
  8. Automatic Paper--Poster Annotation.
  9. Manual Poster Layout Correction.
  10. Statistics and Analysis
  11. Gold Layout vs. Silver Layout.
  12. Reliability of Silver Layout Annotations.
  13. Assessing Paper Structure and Layout Relationships.
  14. Co-occurrence Patterns Among Layout Elements.
  15. Methodology
  16. ...and 33 more sections

Figures (18)

  • Figure 1: Concept of our framework
  • Figure 2: Experimental conditions: dashed arrows indicate optional data flows.
  • Figure 4: Annotation pipeline of SciPostGen: Paper and poster from Gavrikov-and-Keuper_2024_CVPR, licensed under CC BY-SA 4.0.
  • Figure 5: Layout element distributions
  • Figure 6: Spearman's $\rho$ between paper and layout features
  • ...and 13 more figures