← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6606.f · 生成模型:GPT-6

提示链拆分职责也增加接口

理解等级:中级

把复杂任务拆成抽取、校验、格式化等步骤,可以单独定义每步输入输出并定位失败。链路也会增加调用、延迟和中间错误传递;是否优于单次提示需比较实际判据,不能把作者经验当作无条件定理。

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

      Prompt Chaining Often, your tasks need multiple steps and a large number of instructions. One way to go about this is by stuffing all the instructions into a single prompt. An alternative is to break the task into multiple subtasks and chain the prompts such that the output of one prompt determines the input to another. I have observed that prompt chaining consistently performs better

    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:摘录中关于提示链拆分职责也增加接口的定义、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:Prompt Chaining Often, your tasks need multiple steps and a large number of instructions. One way to go about this is by stuffing all the instructions into a single prompt. An alternative is to break the task into multiple subtasks and chain the prompts such that the output of one prompt determines the input to another. I have observed that prompt chaining consistently performs better