ddupont/showdown-distill-llama4

这是一个包含图像和文本的任务型数据集，用于训练模型理解和执行与屏幕截图相关的任务。数据集中的每个示例包含一个屏幕截图、一个问题、一个任务类型、一个答案类型、一个答案、图像的宽度和高度、一个唯一标识符、一组可能的选项和一个正确答案。此外，每个示例还包含一个消息列表，提供关于如何与图像和文本进行交互的上下文信息。

This is a task-oriented dataset containing images and text, designed for training models to understand and perform tasks related to screen captures. Each example in the dataset includes a screen capture, a question, a task type, an answer type, an answer, the width and height of the image, a unique identifier, a set of possible options, and a correct answer. Additionally, each example contains a message list that provides contextual information on how to interact with the images and text.