MathVista, WeMath, ChartQA, LogicVista, MMStar, VisPuzzles, RealWorldQA
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
MathVista, WeMath, ChartQA, LogicVista, MMStar, VisPuzzles, RealWorldQA Unseen Degradations 8B model
64.8MathVista Accuracy
8
MathVista, WeMath, ChartQA, LogicVista, MMStar, VisPuzzles, RealWorldQA Seen Degradations 8B model
73.3MathVista Score
8
MathVista, WeMath, ChartQA, LogicVista, MMStar, VisPuzzles, RealWorldQA 8B model (Clean)
78.5MathVista Accuracy
8