A STAGE REVIEW OF INSTRUCTION TUNING

Summary from the TLDR AI newsletter

Instruction tuning is good at getting models to do what we want. People misunderstand the importance of FLAN because it gives short responses and humans typically prefer long responses. This article also shows other models and when you should consider using them.