← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8694.a · 生成模型:GPT-6

反思要由可识别故障触发

理解等级:中级

重复动作或验证失败可以触发重新检查计划和假设。无条件频繁反思可能让正确答案被改坏;应设置触发条件、修订目标和轮次预算,并以外部证据判断是否改善。

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

      Reflection The agent loop may include self-verification or correction steps. This was pioneered by Shinn et al. with the Reflexion paradigm. Here is the system prompt for Reflection-Llama-3.1 that uses reflection techniques: Prompt: You are a world-class AI system, capable of complex reasoning and reflection. Reason through the query inside <thinking> tags, and then provide your final response inside <output> tags. If you detect that

    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:摘录中关于反思要由可识别故障触发的定义、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:Reflection The agent loop may include self-verification or correction steps. This was pioneered by Shinn et al. with the Reflexion paradigm. Here is the system prompt for Reflection-Llama-3.1 that uses reflection techniques: Prompt: You are a world-class AI system, capable of complex reasoning and reflection. Reason through the query inside <thinking> tags, and then provide your final response inside <output> tags. If you detect that