Dataset · AI4Math
MathVerse: visual math problems
AI4Math/MathVerse
Visual math problem benchmark for multimodal LLMs.
What's inside
The set contains images and text-based math questions in English. Problems are presented in multiple-choice, question-answering, and visual-question-answering formats. The card describes it as a benchmark to test whether MLLMs actually interpret diagrams rather than just text.
What you can do with it
- Evaluate MLLMs on visual math
- Test VQA models on diagrams
- Compare multimodal LLMs on the benchmark
- Tasks
- Multiple choice, Question answering, Visual question answering
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 9.3K
- Likes
- 73
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("AI4Math/MathVerse")