知识卡片
问答改写与回答生成职责不同
多轮检索可以先把当前问句与历史转成独立查询,再检索证据,最后生成回答。改写需保留用户原意与约束;如果改写错误,后续检索和生成即使各自运行正常也可能回答错问题。
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
- · 大语言模型入门
原书出处 · 1 处
- · 查看原书依据
Takes the previous conversational history, if it exists, and the current response/query from the user and creates a standalone question. Uses a chosen retrieval method to retrieve top-k most similar chunks to the question. Takes the retrieved chunks, the conversational history, the current user/response query, and instructions and feeds it to the LLM. The LLM generates the answer. We can call the chain and append
参考来源
- 位置:《Designing Large Language Model Applications》第01章《大语言模型入门》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch01.html)
- 结论依据:摘录中关于问答改写与回答生成职责不同的定义、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:Takes the previous conversational history, if it exists, and the current response/query from the user and creates a standalone question. Uses a chosen retrieval method to retrieve top-k most similar chunks to the question. Takes the retrieved chunks, the conversational history, the current user/response query, and instructions and feeds it to the LLM. The LLM generates the answer. We can call the chain and append