happy8825/MMLongBench_var5_deterministic
收藏Hugging Face2025-12-14 更新2025-12-20 收录
下载链接：
https://hf-mirror.com/datasets/happy8825/MMLongBench_var5_deterministic
下载链接
链接失效反馈官方服务：
资源简介：
MMLongBench是一个多模态长文本问答基准数据集，包含来自不同证据源（图表、纯文本、表格、布局文本等）的样本。数据集结构复杂，支持多轮交互处理，包含文档ID、问题、答案、相关页面、证据来源等特征。评估指标显示在不同证据类型和页面长度下的准确率表现，适用于测试多模态系统的长文本理解能力。
MMLongBench is a multimodal long-form question answering benchmark dataset containing samples from various evidence sources (figures, plain-text, tables, charts, layout-text, etc.). The dataset features complex structures supporting multi-turn interactions, including document IDs, questions, answers, relevant pages, evidence sources and other features. Evaluation metrics show accuracy performance across different evidence types and page lengths, suitable for testing multimodal systems long-text understanding capabilities.
提供机构：
happy8825
