GODBench

该数据集名为GODBench，是一个融合视频和文本模态的综合基准测试，旨在系统地评估多模态大型语言模型在创作评论艺术方面的能力。GODBench提供了一种结构化的方法，用以评估大型多模态语言模型在为视频生成评论时的创造力，同时突显它们在理解和创造艺术内容方面的优势和局限性。该数据集包含了不同规模配置的评论，范围为从1到m、n的配置，其任务是对大型多模态语言模型在创造性地生成视频评论方面的评估。

This dataset, named GODBench, is a comprehensive benchmark integrating video and text modalities, designed to systematically evaluate the capabilities of multimodal large language models in the field of artistic comment creation. GODBench provides a structured methodology to assess how effectively large multimodal language models generate comments for videos, while highlighting their respective strengths and limitations in understanding and creating artistic content. This dataset includes comment configurations with varying scales, ranging from 1 to m and 1 to n, and its core task is to evaluate large multimodal language models on their ability to creatively generate video comments.