知识卡片

普通读书笔记卡 · 6640.d · 生成模型:GPT-6

偏好排序受样本和展示影响

理解等级:中级

匿名成对比较仍会受长度、风格、位置与投票人群影响。可随机化展示、分开评估事实和语言质量,并报告样本与不确定性;总体偏好胜率不等于事实准确率或实际任务效用。

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

      Humans tend to prefer longer texts. Humans tend to overlook subtle factuality and consistency issues if the style is authoritative or convincing. Humans can be indecisive and tend to grant ties instead of choosing a winner. The order in which the LLM answers are presented can influence human ratings. This can be rectified by providing randomized answers to the user. Wu et al. propose a

    参考来源

    - 位置:《Designing Large Language Model Applications》第05章《按任务适配模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch05.html) - 结论依据:摘录中关于偏好排序受样本和展示影响的定义、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:Humans tend to prefer longer texts. Humans tend to overlook subtle factuality and consistency issues if the style is authoritative or convincing. Humans can be indecisive and tend to grant ties instead of choosing a winner. The order in which the LLM answers are presented can influence human ratings. This can be rectified by providing randomized answers to the user. Wu et al. propose a