II-Bench

II-Bench是由深圳先进技术研究院和中国科学院联合创建的图像隐含理解基准数据集，旨在评估多模态大型语言模型（MLLMs）的高阶感知能力。该数据集包含1222张来自六个不同领域的图像，如生活、艺术、社会等，涵盖多种图像类型，包括插图、表情包、海报等。数据集的创建过程涉及从多个知名插图网站收集原始图像，并通过严格的数据筛选和标注流程，确保数据的质量和相关性。II-Bench的应用领域主要集中在评估和提升MLLMs在理解复杂图像隐含意义方面的能力，以推动人工智能向更高级的通用智能发展。

II-Bench is an image implicit understanding benchmark dataset jointly developed by the Shenzhen Institute of Advanced Technology and the Chinese Academy of Sciences, designed to evaluate the high-level perceptual capabilities of multimodal large language models (MLLMs). This dataset comprises 1,222 images spanning six distinct domains including daily life, art, and society, covering diverse image types such as illustrations, memes, posters, and more. The construction of II-Bench entails collecting raw images from multiple reputable illustration-related websites, followed by rigorous data filtering and annotation workflows to ensure the quality and relevance of the dataset. The primary application scope of II-Bench focuses on evaluating and enhancing the ability of MLLMs to comprehend the implicit semantics of complex images, thereby advancing the development of artificial intelligence toward more advanced general artificial intelligence (AGI).