ONE-Lab/LiveVQA-new
收藏Hugging Face2025-07-07 更新2025-10-18 收录
下载链接：
https://hf-mirror.com/datasets/ONE-Lab/LiveVQA-new
下载链接
链接失效反馈官方服务：
资源简介：
LIVEVQA是一个用于评估多模态大型语言模型在处理最新现实世界视觉信息上的能力的基准数据集。它包含了超过107,000个样本，每个样本都是一对图像和经过仔细验证的多层次问题答案集，图像来源于新闻、YouTube或学术论文等最新事件或出版物。LIVEVQA支持对最先进的多模态大型语言模型进行视觉推理和知识更新的研究。每个条目包括一个来自最新事件或出版物的视觉场景、多层次推理问题和答案、用于可复现性的来源和元数据。
LIVEVQA is a benchmark dataset of over 107,000 samples designed to evaluate Multimodal Large Language Models (MLLMs) on up-to-date real-world visual information. Each sample pairs a recent image from sources like news, YouTube, or academic papers, with multi-level, carefully validated question-answer sets. LIVEVQA supports research on visual reasoning and knowledge updating for state-of-the-art MLLMs. Each entry includes a visual scene from a recent event or publication, multi-level reasoning questions with answers, and source and metadata for reproducibility.
提供机构：
ONE-Lab
