niavg/DeepEyesGrounded

这是一个包含rationale、数据来源、提示内容、图片、能力、环境名称、奖励模型和额外信息的复杂结构数据集。数据集中的提示内容包括内容和角色信息，图片字段包含图片的字节和路径信息。奖励模型包括真实情况和风格，额外信息包括答案、ID、索引、预测信息和问题等信息。数据集分为训练集，包含25000个示例。

This is a complex structured dataset containing rationale, data source, prompt content, images, ability, environment name, reward model, and extra information. The prompt content includes information about the content and role. The image field contains the bytes and path of the image. The reward model includes ground truth and style. Extra information includes answer, ID, index, prediction information, and question. The dataset is split into a training set with 25,000 examples.