知识卡片

普通读书笔记卡 · 6606.a · 生成模型:GPT-6

乱序可读不代表顺序不重要

理解等级:中级

模型能从某些打乱词序的句子恢复意图,只说明在这些样本上具有容错性。交换主体、宾语、否定范围或指令顺序仍可能改变含义与输出,不能推广为模型对词序不敏感或概率分布不变。

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

      insensitive to word order. This property has been observed even in earlier models like BERT. For example, ask ChatGPT or your favorite LLM provider the question “How do I tie my shoelaces?” in jumbled form, say “shoe tie my I how do laces?” ChatGPT responds with “Certainly! Here are step-by-step instructions on how to tie your shoelaces:…​” as if you asked a straightforward question.

    参考来源

    - 位置:《Designing Large Language Model Applications》第01章《大语言模型入门》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch01.html) - 结论依据:摘录中关于乱序可读不代表顺序不重要的定义、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:insensitive to word order. This property has been observed even in earlier models like BERT. For example, ask ChatGPT or your favorite LLM provider the question “How do I tie my shoelaces?” in jumbled form, say “shoe tie my I how do laces?” ChatGPT responds with “Certainly! Here are step-by-step instructions on how to tie your shoelaces:…​” as if you asked a straightforward question.