"GraphVQA\u201111K: A Visual Question Answering Dataset on Graphs"

"                \ud83d\udcca GraphVQA\u201111K: Visual Question Answering on Graphs\ud83e\udde9 Dataset DescriptionThe GraphVQA\u201111K dataset is a benchmark for evaluating Visual Question Answering (VQA) models on graph\u2011structured data. Using the networkx library, 11,000 graph images were generated across diverse graph families, each paired with one automatically constructed question\u2013answer (QA) pair. This dataset enables testing of Vision\u2011Language Models (VLMs) on tasks requiring both visual reasoning and graph theory knowledge.It provides a controlled environment for studying multi\u2011modal understanding where visual topology meets textual semantics.\r\nResearchers can use it to explore graph property prediction, visual reasoning accuracy, and cross\u2011domain generalization in VLMs.\r\nThe dataset also serves as a foundation for developing explainable AI frameworks that interpret graph\u2011based visual questions.\ud83c\udf10 Graph TypesThe dataset includes multiple graph families:\ud83d\udd39 Random graphs\ud83d\udd38 Cycle graphs\ud83d\udd39 Path graphs\ud83c\udf1f Star graphs\u26aa Complete graphs\ud83c\udf00 Wheel graphs\ud83e\uddf1 Grid graphs\ud83c\udf32 Tree graphs\ud83d\udd00 Directed random graphs "