I am working on something I am calling "compact models" (as part of a PhD): techniques to pack more information into Machine Learning models when their size is constrained in some way. I have put up some of our work here [1] - it has been interesting so far, and the results are promising. I would like to release a Python library soon, well, ...in a few months - my PhD is part-time, I have a full time day job and time management is a pain.
Interesting name. :) I worked on developing compact models for the first few years of my employment at NXP Semiconductors, which is (for me, at least) the top result of a Google search for "compact models", defining them as "mathematical descriptions of semiconductor devices used in analog circuit simulators".
You're right, thats what the term seems to stand for (I had Googled too, when I was thinking of names). "Compact models" is something I am considering calling this work - its by no means standard :)
[1] https://arxiv.org/abs/1905.01520