知识卡片
可选阶段要有独立增益证据
改写、检索、重排、精炼、插入与生成可组成多阶段管线,验证可置于各阶段。除基本取证与生成外,不应因架构图包含就全部堆上;每项增加成本,需要通过阶段和端到端指标证明价值。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 检索增强生成
原书出处 · 1 处
- · 查看原书依据
The RAG Pipeline A typical RAG application follows the retrieve-read framework, as discussed in Chapter 11. In response to a query, a retrieval model identifies documents that are relevant to answering the query. These documents are then passed along to the LLM as context, which the LLM can rely on in addition to its internal capabilities to generate a response. In practice, we typically need
参考来源
- 位置:《Designing Large Language Model Applications》第12章《检索增强生成》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch12.html)
- 结论依据:摘录中关于可选阶段要有独立增益证据的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:The RAG Pipeline A typical RAG application follows the retrieve-read framework, as discussed in Chapter 11. In response to a query, a retrieval model identifies documents that are relevant to answering the query. These documents are then passed along to the LLM as context, which the LLM can rely on in addition to its internal capabilities to generate a response. In practice, we typically need