TIGER-Lab/PixelWorld

PixelWorld是一个统一了文本、表格、代码、图表和图像为基于像素输入的多模态基准数据集，它能够对基于令牌和基于像素的处理方式进行直接比较，并广泛覆盖了文本、结构化和多模态任务，支持将文本和表格转换为图像以及并行评估。

PixelWorld is a multimodal benchmark that unifies text, tables, code, diagrams, and images into pixel-based inputs, allowing direct comparison between token-based and pixel-based processing, covering a wide range of text-only, structured, and multimodal tasks, supporting conversion of text and tables into images, and parallel evaluation.