Distill or Annotate? How to Most Efficiently Use a Fixed LLM Budget -Extensive experiments on 6 diverse NLP tasks -Distilling LLM (11B) to small language model (60M) is more cost-effective -Compared to annotating more data to directly train small model https://t.co/9lJ09OrcAU https://t.co/c38b9z41IR
— John Nay (@johnjnay) May 3, 2023
from Twitter https://twitter.com/johnjnay
May 02, 2023 at 06:10PM
via IFTTT
No comments:
Post a Comment