FractalGPT/RRNCBFinalPublic

该数据集是RRNCB基准测试的样本，用于评估RAG（检索增强生成）解决方案在处理俄罗斯规范性、法律和技术文档时的性能。数据集包含900个问题和65个PDF文件，分为技术文档（如GOST和SNiP标准）和法律文档（如俄罗斯法典）两大主题。数据集的结构包括问题、答案和文档来源三个字段。该基准测试旨在为AI工程师、数据科学家和商业用户提供一个统一的评估标准，以测试和比较不同RAG解决方案的效果。

This dataset is a sample for the RRNCB benchmark, designed to evaluate the performance of RAG (Retrieval Augmented Generation) solutions in handling Russian normative, legal, and technical documents. The dataset includes 900 questions and 65 PDF files, divided into two main themes: technical documentation (e.g., GOST and SNiP standards) and legal documents (e.g., Russian codes). The dataset structure consists of three fields: question, answer, and document source. The benchmark aims to provide a unified evaluation standard for AI engineers, data scientists, and business users to test and compare the effectiveness of different RAG solutions.