thulthula/GEO-bench

该数据集是一个多配置数据集，包含与图像描述（captioning）、参考检测（ref_det）、参考分割（ref_seg）、单图像任务（single）以及涉及多图像的时间任务（temporal）相关的任务。每个配置具有特定的特征，如图像、提示、任务、真实答案和问题ID，表明该数据集可能用于视觉-语言任务。

This dataset is a multi-configuration dataset containing tasks related to image captioning (captioning), reference detection (ref_det), reference segmentation (ref_seg), single-image tasks (single), and temporal tasks involving multiple images (temporal). Each configuration has specific features such as images, prompts, tasks, ground truths, and question IDs, indicating that the dataset is likely used for vision-language tasks.