← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6756 · 生成模型:GPT-6

多标签评估需报告每类与总体

理解等级:中级

总体指标可能掩盖威胁等稀少标签的漏判,应看各标签分布及精确率、召回等适用指标。单条攻击例不能推断对所有敏感用户已有效,标签主观性也不使所有评价无意义;边界和错误代价仍须明确。

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

      Tip Typically, at this point, you would explore and analyze the data, focusing on the qualities of the text samples and the accuracy of the labels and, perhaps, ask yourself questions about the data. How long are the comments, in general? Does sentence length or comment length have any relation to toxicity? Consider focusing on some of the severe_toxic comments. What sets them apart from

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第09章《可堆叠的Transformer》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-9.html) - 结论依据:摘录中关于多标签评估需报告每类与总体的定义、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:Tip Typically, at this point, you would explore and analyze the data, focusing on the qualities of the text samples and the accuracy of the labels and, perhaps, ask yourself questions about the data. How long are the comments, in general? Does sentence length or comment length have any relation to toxicity? Consider focusing on some of the severe_toxic comments. What sets them apart from