MVBench

该数据集为评估多模态视觉-语言模型提供了一个基准，它专注于涵盖多种多样的多模态任务，其核心任务是进行多模态评估。

This dataset, which serves as a benchmark for evaluating multimodal vision-language models, covers a diverse array of multimodal tasks, with its core task being multimodal evaluation.