RAP Dataset for Paper-to-Poster Visual Coding Generation
October 15, 2026
The RAP dataset introduces 25K+ verified examples for transforming research papers into conference posters via declarative markup. It treats poster generation as a visual coding task to enable objective render-and-compare evaluation instead of subjective VLM-as-judge metrics.
HOW THIS AFFECTS YOU
●
builderYou can use this dataset to train models on complex, multi-step layout and summarization tasks.
●
designerThis targets the automation of highly structured, information-dense visual layouts.