知识卡片
工具标记需要系统执行语义
工具调用和结果分隔符可以标识结构,但实际解析、权限、执行及结果回传由应用协议承担。模型生成一个工具标记不会自动调用API,也不能证明参数正确或操作已完成。
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
- · 词表与分词
原书出处 · 1 处
- · 查看原书依据
<TOOL_CALL>, </TOOL_CALL> Content between these tokens is used as input to an external tool, like an API call or a query to a database. <TOOL_RESULT>, </TOOL_RESULT> Content between these tokens is used to represent the results from calling the aforementioned tools. As we have seen, if our data is domain-specific like healthcare, scientific literature, etc., tokenization from a general-purpose tokenizer will be unsatisfactory. GALACTICA by
参考来源
- 位置:《Designing Large Language Model Applications》第03章《词表与分词》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch03.html)
- 结论依据:摘录中关于工具标记需要系统执行语义的定义、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:<TOOL_CALL>, </TOOL_CALL> Content between these tokens is used as input to an external tool, like an API call or a query to a database. <TOOL_RESULT>, </TOOL_RESULT> Content between these tokens is used to represent the results from calling the aforementioned tools. As we have seen, if our data is domain-specific like healthcare, scientific literature, etc., tokenization from a general-purpose tokenizer will be unsatisfactory. GALACTICA by