opencompass/Creation-MMBench

Creation-MMBench是一个专门为评估多模态大型语言模型（MLLM）创造力而设计的多模态基准数据集。它包含765个测试用例，跨越51个细粒度任务，为MLLM提供了图像和上下文信息，包括角色、背景信息和指令等。数据集还包含了为每个测试用例精心设计的实例特定标准，以评估模型生成的内容的通用响应质量和视觉事实对齐性。此外，该数据集还提供了丰富的创造性问题，并采用多图像格式，每个问题都设计有特定角色，以激发MLLM的创造力。

Creation-MMBench is a multimodal benchmark specifically designed to evaluate the creative capabilities of Multimodal Large Language Models (MLLMs). It features 765 test cases spanning 51 fine-grained tasks, providing MLLMs with images and context, including role, background information, and instructions. The dataset includes carefully crafted instance-specific criteria for each test case to assess both general response quality and visual-factual alignment in model-generated content. Additionally, the dataset offers a rich set of creative questions and adopts a multi-image format, with each question designed to stimulate MLLMs creative capabilities through specific roles.