Say we make a couple language models that represent different viewpoints. And they will have these kinds of gaps.
j 下一段next speechk 上一段previous speech
