Brief

Researchers have developed DiagramRAG, a new framework designed to assist in the generation of scientific diagrams from user sketches. This system leverages retrieval-augmented generation, using knowledge graphs to represent diagrams and an embedding model to find reference diagrams that are both semantically relevant and topologically compatible with the input sketch. The retrieved references then inform the completion and rendering of publication-quality diagrams, achieving strong performance on benchmarks like DiagramBank and FigureBench. AI

Researchers have developed DiagramBank, a new dataset containing over 57,000 schematic diagrams extracted from AI and ML papers hosted on OpenReview. This dataset meticulously links each diagram to its source paper's title, abstract, caption, and in-text references, providing valuable context. DiagramBank is designed to support advancements in scientific document understanding, diagram retrieval, and the creation of new benchmarks, with a reported precision of 93.67% based on a manual audit. AI