知识卡片

普通读书笔记卡 · 9107 · 生成模型:GPT-6

诚实呈现反例能积累可信度

理解等级:中级

若方法在重要条件失败,应明确报告,即使暂时解释不通。反直觉结果可能后来揭示模型前提错误;压掉失败例会让其他人和未来自己失去线索。准确描述局限比只展示成功案例更支持成果复用。

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

      You should develop a reputation for being clear and reliable. Always convey an accurate impression of the performance of an algorithm. If something doesn't work in important cases, let the readers know. In [125], we noted that our deblurring algorithm didn't work for a case where we had been sure it should work for a maximum a posteriori (MAP) estimation of the deblurred image. Some

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第53章《如何写论文》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c053.xhtml) - 结论依据:摘录中关于诚实呈现反例能积累可信度的定义、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:You should develop a reputation for being clear and reliable. Always convey an accurate impression of the performance of an algorithm. If something doesn't work in important cases, let the readers know. In [125], we noted that our deblurring algorithm didn't work for a case where we had been sure it should work for a maximum a posteriori (MAP) estimation of the deblurred image. Some