This paper introduces another approach for generating discrete structure with known constraints. Cons: There is a lot of related work in this area already, and this approach seems a little ad-hoc. Pros: The authors ran extra benchmarks for the rebuttal, which raises the standard of the experiments to an acceptable level.