知识卡片
复述中间材料仍是模型生成
先让模型生成与问题有关的背景段落,再基于段落回答,尝试使过程更接近预训练式续写。背景不是实际检索到的文献,不能把它伪装成来源;多段复述与投票仍需检查共同事实错误。
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
- · 对齐训练与推理
原书出处 · 1 处
- · 查看原书依据
Recitation With the recitation technique, we prompt the LLM to generate one or more passages about the given question and then generate the answer based on the passages generated. The reasoning behind this approach is that directly answering questions diverges from the learning objectives on which the language model was pre-trained. Recitation serves as the intermediate step that aligns more closely to the original learning
参考来源
- 位置:《Designing Large Language Model Applications》第08章《对齐训练与推理》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch08.html)
- 结论依据:摘录中关于复述中间材料仍是模型生成的定义、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:Recitation With the recitation technique, we prompt the LLM to generate one or more passages about the given question and then generate the answer based on the passages generated. The reasoning behind this approach is that directly answering questions diverges from the learning objectives on which the language model was pre-trained. Recitation serves as the intermediate step that aligns more closely to the original learning