知识卡片
往返重建一致不保证内容真实
将输出再还原输入并比较,可检验信息保持的某个代理。两个方向可能共享同一错误或丢掉同类细节,不能把往返相似当原输出已全面准确;还需对应源信息逐项验收。
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
- · 模型自动化测试
原书出处 · 1 处
- · 查看原书依据
round-trip” results to the original inputs. This remains subject to some of the same weirdness issues, however. When it’s not clear what to test or the effort to implement a test is prohibitive, consider generating a “gallery” of representative outputs. This will allow a human expert who “knows it when they see it” to detect failures and also, in time, to more clearly define failure
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:摘录中关于往返重建一致不保证内容真实的定义、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:round-trip” results to the original inputs. This remains subject to some of the same weirdness issues, however. When it’s not clear what to test or the effort to implement a test is prohibitive, consider generating a “gallery” of representative outputs. This will allow a human expert who “knows it when they see it” to detect failures and also, in time, to more clearly define failure