four. The pre-experienced model can act as a great starting point allowing great-tuning to converge speedier than instruction from scratch.