知识卡片
注意力权重不等于因果边
注意力可帮助上下文表示和信息聚合,但注意力较大不证明某token是现实事件的原因。层深与语义粒度也不是固定逐层配方;因果表示主张还需要明确对象、干预语义和独立的识别或实验依据。
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
- · 因果与大语言模型
原书出处 · 1 处
- · 查看原书依据
13.2.2 The attention mechanism One of the main drivers behind the success of LLMs is use of Transformer architectures and other neural network architectures that rely on a mechanism called attention. The attention mechanism allows the model to weigh the importance of different parts of an input sequence differently. That allows the model to learn to “focus” on specific parts of a sequence that are
参考来源
- 位置:《Causal AI》第13章《因果与大语言模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-13.xhtml)
- 结论依据:摘录中关于注意力权重不等于因果边的定义、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:13.2.2 The attention mechanism One of the main drivers behind the success of LLMs is use of Transformer architectures and other neural network architectures that rely on a mechanism called attention. The attention mechanism allows the model to weigh the importance of different parts of an input sequence differently. That allows the model to learn to “focus” on specific parts of a sequence that are