知识卡片
检索器与阅读器分工验收
检索器选择候选文本,阅读器基于文本生成回答。两者应分别检查召回相关证据与忠实使用证据的能力;高向量相似分不代表答案已找到,也不证明模型正确理解证据。
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
- · 表示学习与嵌入
原书出处 · 1 处
- · 查看原书依据
are provided as a response to the search query. This process is called retrieval. The texts are then fed into the LLM prompt along with the user query, and the LLM uses the information provided in the context to answer the user query. This two-step process has traditionally been called the retriever-reader framework, with the LLM playing the role of the reader in this example.
参考来源
- 位置:《Designing Large Language Model Applications》第11章《表示学习与嵌入》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch11.html)
- 结论依据:摘录中关于检索器与阅读器分工验收的定义、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:are provided as a response to the search query. This process is called retrieval. The texts are then fed into the LLM prompt along with the user query, and the LLM uses the information provided in the context to answer the user query. This two-step process has traditionally been called the retriever-reader framework, with the LLM playing the role of the reader in this example.