berhaan/pisc-tr

LLaVA-CoT数据集是一个用于视觉问答任务的多语言数据集，包含英语和土耳其语。数据集的结构包括图像文件和`train.jsonl`文件，后者包含了问答数据，格式为JSON。数据集是从多个开源数据集中创建的，主要用于视觉问答任务。数据集的图像和问题来源于开放数据集，答案由GPT-4生成。尽管已经尽力确保数据的多样性和准确性，但仍可能存在一些偏见。如果发现任何侵权或不适当的内容，请联系数据集维护者进行处理。

The LLaVA-CoT dataset is primarily used for visual question answering tasks, supporting English and Turkish. The dataset includes images and question-answering data, with a structure that includes id, image path, and conversations (including Turkish and English translations). The dataset is created from various open-source datasets, primarily containing 1k entries from the PISC dataset. The creation process of the dataset involved using GPT-4o to generate answers, and efforts were made to ensure data diversity and accuracy, although biases and risks may still exist.