Their paper [which I enjoyed very much] presents some nice pop-gen insights into expected patterns of variation under reciprocal introgression, and raises some interesting Qs about post-admixture selection in humans and Neanderthals.

Read it here: www.cell.com/current-biol...

So for K=1, it's trivial ({} and {{1}}). For K=2: {}, {{1}}, {{2}}, {{1, 2}}, and {{1}, {2}, {1, 2}}. I thought I had a nice iterative method worked out, but it starts to blow up around K=5 checking all the duplicated solutions, even though the final number is not that huge.