原书出处
Designing Large Language Model Applications
input-first approach, where the LLM is asked to generate an input instance for the given instruction and subsequently asked to generate the output label for that input. However, this approach might lead to label imbalance as shown in Wang et al., with certain labels being overrepresented. Therefore, it is a good approach to mix output-first generation, where you ask the LLM to generate the output
原书定位详情
来源:a19c7933-4eda-4176-aebd-30fe7e1ad292
正文位置:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch06.html
EPUB SHA-256:1b4d2347fec2f949fcddfbaf4fad597a1ae20887a096cfae64c59a619b4781bf