Certainly I may be missing something, but it seems like the advance in this series of papers is that they figured out a way to calculate a differentiable solution to the sorting problem quickly, whereas it was already known that the a differentiable solution already existed, no?