知识卡片
隐私检测与处置分开验收
检测决定哪些片段可能敏感,处置决定替换、掩码或删除什么。漏检会让内容未经处理,误检又可能损坏训练信号;两阶段应分别记录召回、误删及对下游任务的影响,而非只统计最终删除数量。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 预训练数据
原书出处 · 1 处
- · 查看原书依据
PII remediation. The Privacy working group (of which I was the co-lead) of the Big Science project that trained the BLOOM model developed a pipeline for PII detection and remediation, which we will discuss next.
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:摘录中关于隐私检测与处置分开验收的定义、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:PII remediation. The Privacy working group (of which I was the co-lead) of the Big Science project that trained the BLOOM model developed a pipeline for PII detection and remediation, which we will discuss next.