NayanaBench is a comprehensive multilingual visual question answering (VQA) dataset designed for evaluating multimodal AI systems across diverse languages and scripts. The dataset combines optical character recognition (OCR), layout analysis, and visual question answering capabilities across 22 languages.

🌍 Languages Supported

The dataset includes 200 examples each for the following languages:

Indian Languages

Kannada (kn) - ಕನ್ನಡ Hindi (hi) -… See the full description on the dataset page: https://huggingface.co/datasets/Nayana-cognitivelab/NayanaBench.