知识卡片
假想文档是检索代理不是证据
HyDE等让模型生成可能回答问题的文档,再用其表示寻找真实文档。虚构人名或事实可以仅作为语言代理,必须防止流入最终事实依据;代理的价值由真实检索召回衡量。
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
In recent years, LLM-driven query expansion methods are gaining more prominence. Two such examples are Query2doc and Hypothetical Document Embeddings (HyDE). Let’s explore how LLM-based query expansion works. We pass a given query to an LLM and ask it to generate a hypothetical document that contains the answer to the query. Prompt: For the provided query, generate a hypothetical document in 2 sentences or less
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:摘录中关于假想文档是检索代理不是证据的定义、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:In recent years, LLM-driven query expansion methods are gaining more prominence. Two such examples are Query2doc and Hypothetical Document Embeddings (HyDE). Let’s explore how LLM-based query expansion works. We pass a given query to an LLM and ask it to generate a hypothetical document that contains the answer to the query. Prompt: For the provided query, generate a hypothetical document in 2 sentences or less