happy8825/MMLongBench_var8_doc_memory
收藏Hugging Face2025-12-18 更新2025-12-20 收录
下载链接：
https://hf-mirror.com/datasets/happy8825/MMLongBench_var8_doc_memory
下载链接
链接失效反馈官方服务：
资源简介：
MMLongBench数据集是一个用于评估模型性能的数据集，包含多种证据来源（如图表、纯文本、布局文本和表格）和不同长度的证据页面。数据集配置为var8_doc_memory，包含10个训练样本，每个样本具有多个特征，如相关页面、证据页面、分数、文档ID、问题、答案等。性能指标包括平均准确率和按证据来源及页面长度分类的子集指标。
The MMLongBench dataset is designed for evaluating model performance, incorporating various evidence sources (e.g., charts, plain text, layout text, and tables) and evidence pages of different lengths. Configured as var8_doc_memory, the dataset includes 10 training samples, each with multiple features such as relevant pages, evidence pages, scores, document IDs, questions, answers, etc. Performance metrics include average accuracy and subset metrics categorized by evidence source and page length.
提供机构：
happy8825
