Measuring Multimodal Mathematical Reasoning with MATH-Vision Dataset

Benchmark Model Rank Results
multimodal-reasoning-on-math-vGPT4V#1Accuracy: 22.76
multimodal-reasoning-on-math-vGemini Pro#2Accuracy: 17.66
multimodal-reasoning-on-math-vQwen-VL-Max#3Accuracy: 15.59
multimodal-reasoning-on-math-vInternLM-XComposer2-VL#4Accuracy: 14.54