https://www.quora.com/Why-is-using-a-GUI-in-most-of-all-prog...
In that answer I link to the 7 GUIs project from Eugen Kiss: https://github.com/eugenkiss/7guis which I think tried to answer your question from an evaluative, example-based approach. What I like about it was that it kept the requirements simple, but moved beyond Hello World, so it's a bit more fair to systems that have a higher lower bound for minimal GUIs.
For the last part of your question, think in terms of generality vs a DSL-based approach. A GUI system that is completely general, and can't make any assumptions, is naturally going to require more information for a given output. Once you narrow down the possibilities, and include defaults and the ability to configure things for re-use (e.g. widget styles and themes), you can dramatically reduce the effort needed. Constraint based systems (e.g. Cassowary) are another way to go.