知识卡片
微调与检索改变不同层面
检索按需提供证据,微调改变模型处理任务、领域及输出风格的行为。两者可组合,也可微调检索和摘要器;某知识问答实验RAG占优不代表所有任务都不需微调。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 检索增强生成
原书出处 · 1 处
- · 查看原书依据
RAG Versus Fine-Tuning The debate around using RAG versus fine-tuning boils down to the more fundamental question: what aspects of the task can I perform using the LLM versus relying on external sources? In cases where external knowledge is required to solve a task, both retrieval and fine-tuning can be used. Retrieval can be used to integrate the knowledge on demand, with the drawback being
参考来源
- 位置:《Designing Large Language Model Applications》第12章《检索增强生成》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch12.html)
- 结论依据:摘录中关于微调与检索改变不同层面的定义、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:RAG Versus Fine-Tuning The debate around using RAG versus fine-tuning boils down to the more fundamental question: what aspects of the task can I perform using the LLM versus relying on external sources? In cases where external knowledge is required to solve a task, both retrieval and fine-tuning can be used. Retrieval can be used to integrate the knowledge on demand, with the drawback being