DesignProbe

DesignProbe是一个专为评估多模态大型语言模型（MLLMs）在图形设计领域能力而建立的基准。该数据集涵盖了八个任务，分为元素级和整体设计级两个层次。元素级任务关注颜色、字体和布局等设计元素的属性识别和语义理解；整体设计级任务则涉及风格分类和视觉隐喻的理解。DesignProbe通过这些任务，旨在深入探索MLLMs在设计领域的理解和应用能力，为解决图形设计中的复杂问题提供新的视角和工具。

DesignProbe is a benchmark specifically established to evaluate the capabilities of multimodal large language models (MLLMs) in the field of graphic design. This dataset covers eight tasks divided into two levels: element-level and holistic design-level. Element-level tasks focus on attribute recognition and semantic understanding of design elements such as color, font, and layout, while holistic design-level tasks involve style classification and comprehension of visual metaphors. Through these tasks, DesignProbe aims to thoroughly explore the understanding and application abilities of MLLMs in the design domain, offering new perspectives and tools for resolving complex issues in graphic design.