tarekmasryo/RAG-QA-Logs-Corpus

RAG QA评估日志与语料库是一个多表数据集，模拟了一个生产风格的检索增强生成(RAG)系统的端到端流程。它包括文档语料库、分块级别索引、每个查询的检索事件、带有正确性、幻觉、延迟和成本信号的QA评估运行、QA用例的场景模板以及记录每一列的数据字典。所有记录都是完全合成但现实的系统日志和语料库数据，旨在模拟真实RAG遥测数据，同时保持安全和可共享性。数据集适用于RAG质量分析、元建模、幻觉风险评分和仪表板式遥测。

The RAG QA Evaluation Logs & Corpus is a multi-table dataset modeling a production-style RAG (Retrieval-Augmented Generation) system end-to-end. It includes a document corpus, a chunk-level index, retrieval events per query, QA evaluation runs with correctness, hallucination, latency, and cost signals, scenario templates for QA use cases, and a data dictionary documenting every column. All records are fully synthetic but realistic system logs and corpus data — designed to look and behave like real RAG telemetry while remaining safe to share and experiment with. The dataset is intended for RAG quality analysis, meta-modeling, hallucination risk scoring, and dashboard-style telemetry.