GeoReasoning-10K is a multimodal reasoning dataset in the field of geometry, which was released in 2025 by the University of Illinois at Urbana-Champaign, Shanghai Jiao Tong University, Rutgers University and other institutions. The related paper results are "Generalizable Geometric Image Caption Synthesis", which aims to bridge the gap between visual and linguistic modalities in the field of geometry.

The dataset contains 10,000 geometric image-text pairs with detailed geometric reasoning annotations. Each pair maintains consistency in geometric structure, semantic expression, and visual presentation, forming a highly accurate cross-modal semantic alignment mapping.