知识卡片
行动观察循环应以真实结果推进
ReAct把规划、行动和观察串成循环,帮助交替获取信息和调整步骤。观察必须来自实际工具或环境,不能让模型自行编造;该格式是组织方式,不保证推理或行动正确,也不要求将全部内部推理作为对外交付。
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
- · 外部工具与智能体
原书出处 · 1 处
- · 查看原书依据
I find that a prompt like this is sufficient for most use cases. However, if you feel like the model is not reasoning correctly, you can try ReAct prompting. ReAct At the time of this writing, ReAct (Reasoning + Acting) prompting is the most popular prompt for the agent loop. A typical ReAct prompt looks like this: Prompt: You are an AI assistant capable of
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:摘录中关于行动观察循环应以真实结果推进的定义、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:I find that a prompt like this is sufficient for most use cases. However, if you feel like the model is not reasoning correctly, you can try ReAct prompting. ReAct At the time of this writing, ReAct (Reasoning + Acting) prompting is the most popular prompt for the agent loop. A typical ReAct prompt looks like this: Prompt: You are an AI assistant capable of