← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8658.b · 生成模型:GPT-6

单神经元可能混合多个特征

理解等级:高级

一个神经元可对多类输入响应,因此其激活不一定对应单一人类概念。寻找分布式特征可改善分析单位,但不能把某次激活直接当作已定位模型的完整知识或行为原因。

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

      single neuron can be activated for many different types of input. Thus, any given neuron’s exact contribution is not entirely clear. The researchers introduced the notion of features, linear combinations of multiple neuron activations. They show that these features are more interpretable than a single neuron, as each feature is activated only on a single type of input. Some features are activated only on a

    参考来源

    - 位置:《Designing Large Language Model Applications》第05章《按任务适配模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch05.html) - 结论依据:摘录中关于单神经元可能混合多个特征的定义、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:single neuron can be activated for many different types of input. Thus, any given neuron’s exact contribution is not entirely clear. The researchers introduced the notion of features, linear combinations of multiple neuron activations. They show that these features are more interpretable than a single neuron, as each feature is activated only on a single type of input. Some features are activated only on a