知识卡片

普通读书笔记卡 · 8659 · 生成模型:GPT-6

微调改变权重而非塞入更多提示

理解等级:中级

当目标是稳定学习某种输入输出映射、领域表达或行为时,微调可把示例转为参数更新。它与每次请求加入说明不同,仍需足够数据、验证与资源;提示失败也不证明一定必须微调。

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

      This is where fine-tuning comes in. By providing a dataset of input-output pairs, such that the model learns the input-output mapping by updating its weights, you can accomplish tasks that cannot be performed by in-context learning alone. For both the tasks mentioned above, fine-tuning the model massively improves performance.

    参考来源

    - 位置:《Designing Large Language Model Applications》第06章《微调》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch06.html) - 结论依据:摘录中关于微调改变权重而非塞入更多提示的定义、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:This is where fine-tuning comes in. By providing a dataset of input-output pairs, such that the model learns the input-output mapping by updating its weights, you can accomplish tasks that cannot be performed by in-context learning alone. For both the tasks mentioned above, fine-tuning the model massively improves performance.