when working with language models, you want to be as specific as possible, so "almost the same" unfortunately won't cut it