Something I like to tell it to do is actually to respond using only a certain number of words. Sorta like min token length rather than max token length. Haven'

Something I like to tell it to do is actually to respond using only a certain number of words. Sorta like min token length rather than max token length. Haven't found a model that does it reliably yet.