知识卡片

普通读书笔记卡 · 6610.a · 生成模型:GPT-6

基础与适配模型来自不同阶段

理解等级:中级

基础模型来自预训练;后续指令监督、偏好训练或领域继续预训练,分别改变指令行为、偏好或领域适配。模型名称如chat/instruct只是线索,实际采用的训练阶段与数据仍需核对。

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

      base models. Lately, model providers have been augmenting the base model by fine-tuning it on much smaller datasets to steer them toward being more aligned with human needs and preferences. Some popular tuning modes are:

    参考来源

    - 位置:《Designing Large Language Model Applications》第02章《预训练数据》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch02.html) - 结论依据:摘录中关于基础与适配模型来自不同阶段的定义、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:base models. Lately, model providers have been augmenting the base model by fine-tuning it on much smaller datasets to steer them toward being more aligned with human needs and preferences. Some popular tuning modes are: