🎉 We present FERMAT, a benchmark designed to rigorously test the multimodal reasoning and auto-evaluation capabilities of Vision-Language Models (VLMs) using real-world handwritten math problems! 📝🧠
2,244 handwritten math solutions, carefully annotated across core mathematical domains:
🔢 Arithmetic | 📏 Algebra | 📐 Geometry | 📏 Mensuration | 🎲 Probability | 📊 Statistics | 📐 Trigonometry | 📈… See the full description on the dataset page:
https://huggingface.co/datasets/ai4bharat/FERMAT.
