知识卡片

普通读书笔记卡 · 6640.e · 生成模型:GPT-6

安全评估须写清标准与场景

理解等级:中级

比较模型评估时需问数据是否可见、难度分布如何、与训练是否重合、采用何种安全或帮助性定义。模型按不同标准训练时,单一胜率未必可直接解释为普遍安全程度。

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

      What definition do they use for safety? Can there be a bias in the evaluation due to models being evaluated on the basis of a particular definition of safety, which Llama 2 was trained to adhere to, while other models may have different definitions of safety? Rigorously interrogating the results this way helps you develop a deeper understanding of what is being evaluated, and whether

    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:摘录中关于安全评估须写清标准与场景的定义、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:What definition do they use for safety? Can there be a bias in the evaluation due to models being evaluated on the basis of a particular definition of safety, which Llama 2 was trained to adhere to, while other models may have different definitions of safety? Rigorously interrogating the results this way helps you develop a deeper understanding of what is being evaluated, and whether