知识卡片
生成检索先建立文档标识映射
让模型输出docID而非直接向量匹配,需要文档与标识之间稳定可学的对应关系。标识可用专门token、文本片段、层级簇或关键词构成,唯一性和更新方式决定能否可靠定位。
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
Generative retrieval What if the LLM could identify the right documents(s) that need to be retrieved in response to a query, thus removing the need for retrieval techniques? This is called generative retrieval. Generative retrieval is implemented by assigning identifiers to documents called docIDs, and teaching the LLM the association between documents and docIDs. A document can be associated with one or more docIDs. Typical
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:摘录中关于生成检索先建立文档标识映射的定义、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:Generative retrieval What if the LLM could identify the right documents(s) that need to be retrieved in response to a query, thus removing the need for retrieval techniques? This is called generative retrieval. Generative retrieval is implemented by assigning identifiers to documents called docIDs, and teaching the LLM the association between documents and docIDs. A document can be associated with one or more docIDs. Typical