知识卡片

普通读书笔记卡 · 6637 · 生成模型:GPT-6

下一词分数不能直接当解题判断

理解等级:中级

某个答案token未排首位可能受格式、分词、提示与能力共同影响。只看top列表无法证明模型在随机猜题,也不能将一个数字概率当完整解答的正确率;应按任务协议检查最终输出。

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

      correct answer has the 17th highest score. A lot of numbers appear in the top 10, showing that the model is more or less randomly guessing the answer, which is not surprising for a smaller model like GPT Neo.

    参考来源

    - 位置:《Designing Large Language Model Applications》第04章《架构与学习目标》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch04.html) - 结论依据:摘录中关于下一词分数不能直接当解题判断的定义、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:correct answer has the 17th highest score. A lot of numbers appear in the top 10, showing that the model is more or less randomly guessing the answer, which is not surprising for a smaller model like GPT Neo.