← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8693.a · 生成模型:GPT-6

个性化偏好与任务事实分开存

理解等级:中级

用户的单位偏好可跨会话复用,某一天的温度则有时间和地点限制。存储与检索时保留作用范围及来源,避免将旧事实或一次性请求错误推广成长期规则;用户标记满意也不等于外部数据绝对正确。

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

      LLM observation: The user has a preference for Celsius units for temperature data. LLM Output: Let’s perform this step-by-step. First, let’s retrieve today’s temperature. Next, let’s retrieve tomorrow’s temperature. We calculate the difference and generate the answer. External verifier: Task decomposition is correct Tool selection: <WeatherAPI> selected Tool info retrieval: Retrieve information relevant to the query Find temperature data from <WeatherAPI> Tool invocation: get_temperature_curweek(Paris) Tool

    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:摘录中关于个性化偏好与任务事实分开存的定义、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:LLM observation: The user has a preference for Celsius units for temperature data. LLM Output: Let’s perform this step-by-step. First, let’s retrieve today’s temperature. Next, let’s retrieve tomorrow’s temperature. We calculate the difference and generate the answer. External verifier: Task decomposition is correct Tool selection: <WeatherAPI> selected Tool info retrieval: Retrieve information relevant to the query Find temperature data from <WeatherAPI> Tool invocation: get_temperature_curweek(Paris) Tool