知识卡片

普通读书笔记卡 · 6607 · 生成模型:GPT-6

多候选还需要可靠选择标准

理解等级:中级

生成多个候选可以减少单次随机采样失误的影响,但需要有能力筛选正确或适用答案的判据。候选错误可能高度相关,选择器也会出错;额外生成数提高成本,不自动提供可靠性保证。

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

      postprocessing function (which could involve an LLM call) to choose the best one. This is because the LLM samples the generated text from a probability distribution, and in some cases the answer might be wrong/bad just due to an unlucky token sampling. You might have to balance this process against your budget limitations.

    参考来源

    - 位置:《Designing Large Language Model Applications》第01章《大语言模型入门》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch01.html) - 结论依据:摘录中关于多候选还需要可靠选择标准的定义、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:postprocessing function (which could involve an LLM call) to choose the best one. This is because the LLM samples the generated text from a probability distribution, and in some cases the answer might be wrong/bad just due to an unlucky token sampling. You might have to balance this process against your budget limitations.