知识卡片

普通读书笔记卡 · 6517.d · 生成模型:GPT-6

偏好胜率是相对观测

理解等级:中级

成对偏好反映给定题目、展示形式与投票人群下哪个答案更受喜欢。它不等同于绝对正确率,模型名称的排名也依赖实际版本和对手集合;题目选择、风格与用户结构都可能改变观测。

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

      The leaderboard method mentioned in figure F.4 is a judgment-based approach where models are ranked not by accuracy values or other fixed benchmark scores but by user (or other LLM) preferences on their outputs.

    参考来源

    - 位置:《Build a Reasoning Model (From Scratch)》第10章《附录F 常见模型评估方法》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Build a Reasoning Model (From Scratch)/_epub-src/OEBPS/Text/appendix-f.html) - 结论依据:摘录中关于偏好胜率是相对观测的定义、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:The leaderboard method mentioned in figure F.4 is a judgment-based approach where models are ranked not by accuracy values or other fixed benchmark scores but by user (or other LLM) preferences on their outputs.