I think it really depends on what the data DS was fine tuned on. If your use case is very specific, it wouldn’t have distilled that knowledge well.