cmarkea/eval-rag
收藏Hugging Face2025-06-27 更新2025-09-13 收录
下载链接：
https://hf-mirror.com/datasets/cmarkea/eval-rag
下载链接
链接失效反馈官方服务：
资源简介：
这是一个专门设计用于评估基于文档的任务中检索增强生成（RAG）系统质量的多模态评估数据集。每个示例包括一个PDF页面图像、基于页面可见内容自动生成的问答（QA）对、来源类型（文本、表格、信息图、公式或项目列表）以及由多个大型语言模型（LLM）作为评估者提供的人类似判断。
This is a multimodal evaluation dataset specifically designed to assess the quality of Retrieval-Augmented Generation (RAG) systems on document-centric tasks. Each example consists of a PDF page image, an automatically generated question-answer (QA) pair strictly based on the visible content of the page, the source type (text, table, infographic, formula, or bulleted list), and human-like judgments provided by multiple Large Language Models (LLMs) acting as evaluators.
提供机构：
cmarkea
