知识卡片

普通读书笔记卡 · 8667.a · 生成模型:GPT-6

种子质量限定合成数据方向

理解等级:中级

少量清晰、覆盖边界的种子可以指导后续任务生成,但教师可能模仿表面风格或偏离目标分布。应追踪种子来源、生成条件、重复与错误,合成规模扩大后也要保留代表性审查样本。

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 Self-Instruct and Unnatural Instructions papers are the first attempts in this regard. Both start from a seed set of high-quality hand-generated examples, and then in a few-shot setting, ask the LLM to generate similar examples with more diverse linguistic expressions.

    参考来源

    - 位置:《Designing Large Language Model Applications》第06章《微调》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch06.html) - 结论依据:摘录中关于种子质量限定合成数据方向的定义、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 Self-Instruct and Unnatural Instructions papers are the first attempts in this regard. Both start from a seed set of high-quality hand-generated examples, and then in a few-shot setting, ask the LLM to generate similar examples with more diverse linguistic expressions.