知识卡片

普通读书笔记卡 · 8712 · 生成模型:GPT-6

外部记忆扩大可取回范围

理解等级:中级

会话、任务记录和偏好可放外部存储,按需取回有限上下文。它扩大潜在可访问历史,不改变模型一次实际上下文上限,也不保证永不遗忘;关键在于相关记忆是否被正确选择。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      RAG for Memory Management An underrated application of RAG is expanding the context window of an LLM. To recap, an LLM prompt typically contains the following types of (optional) content: The pre-prompt or system prompt These are the overarching instructions provided to the LLM included at the beginning of every query. Depending on your customization needs, the system prompt could occupy a significant part of

    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:摘录中关于外部记忆扩大可取回范围的定义、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:RAG for Memory Management An underrated application of RAG is expanding the context window of an LLM. To recap, an LLM prompt typically contains the following types of (optional) content: The pre-prompt or system prompt These are the overarching instructions provided to the LLM included at the beginning of every query. Depending on your customization needs, the system prompt could occupy a significant part of