General-Level/General-Bench-Closeset

Generalist-Bench是一个大规模的多模态基准数据集，包含超过700个任务和325,800个实例，涵盖图像、视频、音频、3D和语言等多种模态，分为开放和封闭两种基准。开放基准提供完整标注，封闭基准仅提供问题标注。数据集旨在评估模型在多种技能和模态上的表现。

Generalist-Bench is a large-scale multimodal benchmark dataset containing over 700 tasks and 325,800 instances across various modalities including image, video, audio, 3D, and language. It is divided into open and closed benchmarks, with the open benchmark providing full annotations and the closed benchmark only providing question annotations. The dataset is designed to evaluate the performance of models on a wide range of skills and modalities.