MathVista benchmark shows GPT-4V achieves 49.9% accuracy on visual mathematical reasoning tasks, outperforming other models but trailing humans by 10.4%.
" choices = [’3/11’, ’8/11’, ’6/11’, ’3/5’] description =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2023 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts
MathVista benchmark shows GPT-4V achieves 49.9% accuracy on visual mathematical reasoning tasks, outperforming other models but trailing humans by 10.4%.