m-a-p/OmniBench
收藏Hugging Face2025-01-31 更新2025-04-08 收录
下载链接：
https://hf-mirror.com/datasets/m-a-p/OmniBench
下载链接
链接失效反馈官方服务：
资源简介：
OmniBench数据集是一个新颖的基准测试，旨在严格评估模型在视觉、音频和文本输入同时作用下的识别、解释和推理能力。该数据集包含7种任务类型和3种音频类型，以及与问题、答案和选项相关的内容。数据集中的每个样本都包括音频和图像的路径，以及相应的文本内容。
OmniBench is a novel benchmark designed to rigorously evaluate models ability to recognize, interpret, and reason across visual, acoustic, and textual inputs simultaneously. The dataset includes 7 task types and 3 audio types, along with related content for questions, answers, and options. Each sample in the dataset consists of paths to audio and image files, as well as corresponding textual content.
提供机构：
m-a-p
