General-Level/General-Bench-Openset
收藏Hugging Face2025-08-04 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/General-Level/General-Bench-Openset
下载链接
链接失效反馈官方服务：
资源简介：
On Path to Multimodal Generalist: Levels and Benchmarks是一个大规模多模态基准数据集，包含700多个任务和325,800个实例。它涵盖了图像、视频、音频、3D和语言等多种模态，并在各个模态下有详细的任务分类和层次结构。数据集分为开放和封闭两种基准，开放基准提供完整地面真实注释，封闭基准只提供问题注释。
On Path to Multimodal Generalist: Levels and Benchmarks is a large-scale multimodal benchmark dataset consisting of over 700 tasks and 325,800 instances. It covers various modalities including image, video, audio, 3D, and language, each with detailed task taxonomy and hierarchy. The dataset is divided into open and closed benchmarks, with the open benchmark providing full ground-truth annotations and the closed benchmark offering only question annotations without ground-truth answers.
提供机构：
General-Level
