There are many variations to the Card Sort research method, but as usual it depends on the specific question or hypothesis that you are trying to test.
I assume that you want to figure out the groupings for different sections of a form, and therefore you want to how the users might group different types of information.
You'll probably find that card sorting is much more effective for working out taxonomy and hierarchy because you are working with a large number of topics/items and there is more power in the analysis that it provides.
If you want to test the idea of grouping form fields, there are many standard patterns and best practices that you can probably such as carrying out a usability test and provide some variations to compare the completion time and success rate for various tasks.
In terms of using the labels or examples, I think it is common practice that if you use a label then you provide an example so that there is less ambiguity for the user as to how to group them. And if you use examples often you'll have to set a context so that the users have a better idea of how they might actually group the items in that particular scenario.