On the issue of efficiency parameter counting in the paper

Hi, I am a little confused about the details of calculating the number of efficiency parameters in the paper. Not only the method in this paper, but also it seems to be different from that `compactor` paper. Am I missing something? Can you give more calculation details?