Evaluating RAG with LLM as a Judge

summary_zh: Mistral 提出用结构化输出（Structured Outputs）实现 RAG Triad 评估框架，通过 Context Relevance、Groundedness、Answer Relevance 三个指标，结合"LLM As a Judge"对检索与生成结果进行自动评分。