原书出处
Designing Large Language Model Applications
top-p sampling leads to more hallucinations compared to greedy decoding. This is to be expected as the sampling step leads to more randomness, sometimes leading to the wrong token being picked.
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/ch08.html
EPUB SHA-256:1b4d2347fec2f949fcddfbaf4fad597a1ae20887a096cfae64c59a619b4781bf