mm-eval/MMBench
收藏资源简介:
该数据集是一个多模态数据集,包含三个配置:cc、cn和en,每个配置由图像和文本消息组成。特征包括id(唯一标识符)、media(图像列表)和messages(字符串文本)。cc配置仅包含测试集,有2040个样本;cn和en配置均包含开发集(4329个样本)和测试集(6666个样本)。数据集可能用于图像-文本交互任务,如视觉问答或对话生成,但具体应用场景未在README中说明。
This dataset is a multimodal dataset comprising three configurations: cc, cn, and en, each consisting of images and text messages. Features include id (unique identifier), media (list of images), and messages (string text). The cc configuration includes only a test set with 2040 examples; the cn and en configurations both include a dev set (4329 examples) and a test set (6666 examples). The dataset may be intended for image-text interaction tasks, such as visual question answering or dialogue generation, but specific application scenarios are not detailed in the README.




