知识卡片

普通读书笔记卡 · 6630.c · 生成模型:GPT-6

注意力图不能消除指代歧义

理解等级:中级

注意力可帮助建立上下文联系,但示意图偏向某个名字不证明代词在原句中只能指该人。需要额外语境才能消除某些语言歧义;模型内部权重也不能直接当成无歧义的指代标注。

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

      Figure 4-3 depicts how the representation for the token he is heavily weighted by the representation of the token Mark. In this case, the token he is a pronoun used to describe Mark in shorthand. In NLP, mapping a pronoun to its referent is called co-reference resolution.

    参考来源

    - 位置:《Designing Large Language Model Applications》第04章《架构与学习目标》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch04.html) - 结论依据:摘录中关于注意力图不能消除指代歧义的定义、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:Figure 4-3 depicts how the representation for the token he is heavily weighted by the representation of the token Mark. In this case, the token he is a pronoun used to describe Mark in shorthand. In NLP, mapping a pronoun to its referent is called co-reference resolution.