← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8695.b · 生成模型:GPT-6

验证失败需要明确处置策略

理解等级:中级

失败后可重新生成、局部修复、过滤字段、拒绝、抛错或只记录。策略要依据失败类型及可接受后果选择;只记录不会防止错误输出,局部删除或修复也应再次验证依赖字段是否仍一致。

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

      Re-ask In this method, the LLM is asked to regenerate the output, with the prompt containing instructions to specifically abide by the criteria on which the output previously failed validation. Fix In this method, the library fixes the output by itself without asking the LLM for a regeneration. Fixes can involve deletion or replacement of certain parts of the input or output. Filter If structured

    参考来源

    - 位置:《Designing Large Language Model Applications》第10章《外部工具与智能体》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch10.html) - 结论依据:摘录中关于验证失败需要明确处置策略的定义、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:Re-ask In this method, the LLM is asked to regenerate the output, with the prompt containing instructions to specifically abide by the criteria on which the output previously failed validation. Fix In this method, the library fixes the output by itself without asking the LLM for a regeneration. Fixes can involve deletion or replacement of certain parts of the input or output. Filter If structured