知识卡片

普通读书笔记卡 · 6499.c · 生成模型:GPT-6

推理预算应匹配任务难度

理解等级:高级

多步数学和逻辑任务可能受益于更多计算,简单任务则未必需要相同预算。过长推导还可能引入新错误。应以任务验证决定何时增加计算,而不是把某类模型或较长回答普遍视为更好。

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

      Also noteworthy is the mention of knowing "when to think for a long time or not." This hints at an important design consideration: reasoning is not always necessary or desirable.

    参考来源

    - 位置:《Build a Reasoning Model (From Scratch)》第01章《理解推理模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Build a Reasoning Model (From Scratch)/_epub-src/OEBPS/Text/chapter-1.html) - 结论依据:摘录中关于推理预算应匹配任务难度的定义、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:Also noteworthy is the mention of knowing "when to think for a long time or not." This hints at an important design consideration: reasoning is not always necessary or desirable.