The Turkish Visual Reasoning Dataset is a Turkish multimodal reasoning dataset designed to evaluate and improve the abstract reasoning capabilities of Vision-Language Models (VLMs).
It was created by adapting established visual reasoning benchmarks into Turkish and combining them with original Turkish BİLSEM preparation questions. The dataset targets challenging reasoning tasks such as logical pattern discovery, spatial reasoning, analogical… See the full description on the dataset page:
https://huggingface.co/datasets/Berkesule/Turkish-Visual-Reasoning-Dataset.
