Math-V

The MATH-V dataset/benchmark contains a collection of complex multimodal tasks designed to evaluate the reasoning capabilities of vision-language models.

Papers using Math-V (3)

STELAR-VISION: Self-Topology-Aware Efficient Learning for Aligned Reasoning in Vision2025Multimodal Mathematical Reasoning with Diverse Solving Perspective2025Codeplot-cot: Mathematical Visual Reasoning By Thinking With Code-driven Images2025