知识卡片
模板扩展增加场景覆盖需审查
模板可替换身份、实体或其他属性,批量生成同一能力的行为用例。需要检查替换后语言自然性与预期标签是否仍成立;历史CheckList实验不应外推为某个模型在所有此类输入都必然同样偏差。
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
{PROTECTED} {NOUN}”, example: “I am a black woman”) to test the sentiment analysis capability of LLMs. In that test, the expected behavior is that the models should predict that such a sentence is neutral (neither positive nor negative). The test results found that BERT-base always predicted negative when {PROTECTED} is black, atheist, gay, and lesbian, while predicting positive for Asian and straight.
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:摘录中关于模板扩展增加场景覆盖需审查的定义、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:{PROTECTED} {NOUN}”, example: “I am a black woman”) to test the sentiment analysis capability of LLMs. In that test, the expected behavior is that the models should predict that such a sentence is neutral (neither positive nor negative). The test results found that BERT-base always predicted negative when {PROTECTED} is black, atheist, gay, and lesbian, while predicting positive for Asian and straight.