← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8679.a · 生成模型:GPT-6

证据缺字段应保留未知

理解等级:初级

若材料只给人名和奖项,却未给出生日期,不能因前半部分有证据就顺势补齐后半部分。应逐项区分可支持字段、未知字段与需另查来源的字段;部分正确不能掩盖新增事实没有依据。

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

      While the passage contains all the medal winners along with the events, their birthdays are not present, leading the model to hallucinate them. Here is the model’s response: Here is a list of Ugandan 2020 Olympics medal winners with their birthdays: Joshua Cheptegei (born September 12, 1996) - Gold medal in Men’s 5000m Peruth Chemutai (born August 10, 1999) - Gold medal in Women’s 3000m

    参考来源

    - 位置:《Designing Large Language Model Applications》第08章《对齐训练与推理》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch08.html) - 结论依据:摘录中关于证据缺字段应保留未知的定义、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:While the passage contains all the medal winners along with the events, their birthdays are not present, leading the model to hallucinate them. Here is the model’s response: Here is a list of Ugandan 2020 Olympics medal winners with their birthdays: Joshua Cheptegei (born September 12, 1996) - Gold medal in Men’s 5000m Peruth Chemutai (born August 10, 1999) - Gold medal in Women’s 3000m