PosterForest: AI Framework for Scientific Poster Generation

A new framework named PosterForest has been developed by researchers for the automated creation of scientific posters without the need for training, overcoming the shortcomings of current techniques. It employs a structured intermediate format known as the Poster Tree, which encapsulates the hierarchy of documents and the visual-textual semantics at various levels. Agents for content and layout engage in hierarchical reasoning and recursive refinement, enhancing posters from overall structure to detailed composition. This combined optimization leads to better semantic coherence, logical progression, and visual appeal. Experiments indicate that PosterForest surpasses previous methods in both automated and human assessments. The research paper can be found on arXiv with the identifier 2508.21720.