4 ms·
“According To..“ Prompting LLMs Improves Quoting from Pre-Training Data
- oidar 3y agoWhile this isn't a groundbreaking prompting technique, it does make me wonder if we created a training set that only had high quality text, how different would it be that GPT-4.