Figure It Out: Improving the Frontier of Reasoning with Active Visual Thinking

Benchmark Model Rank Results
mathematical-reasoning-on-minerva-mathFigR#19Accuracy: 44.49
mathematical-reasoning-on-math-500FigR#29Accuracy: 95.0
mathematical-reasoning-on-aime24FigR#34Acc: 79.58
mathematical-reasoning-on-aime-2025FigR#42Accuracy: 79.32