知识卡片

普通读书笔记卡 · 8761 · 生成模型:GPT-6

分层指标揭示总体均值盲区

理解等级:中级

总体表现改善可能同时掩盖某国家或职业群体大幅退化。应根据真实应用和潜在伤害选关键切片,检查样本量及统计稳定性;单一切片达标仍不保证交叉群体都达标。

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

      enough. For example, a model might report a high overall performance, but consistently underperform in certain segments of the data. The authors of the seminal “ML Test Score” paper illustrate the usefulness of these tests with an example: The global accuracy of a model may improve by 1%, but accuracy for one country could drop by 50%. This is known as the hidden stratification problem.

    参考来源

    - 位置:《Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)》第06章《模型自动化测试》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)/_epub-src/OEBPS/ch06.html) - 结论依据:摘录中关于分层指标揭示总体均值盲区的定义、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:enough. For example, a model might report a high overall performance, but consistently underperform in certain segments of the data. The authors of the seminal “ML Test Score” paper illustrate the usefulness of these tests with an example: The global accuracy of a model may improve by 1%, but accuracy for one country could drop by 50%. This is known as the hidden stratification problem.