The r is not a bag right? It is a set. I am assuming that [ _cat_n_1<4:7> LBL: h7 ARG0: x3 ] is different from [ _cat_n_1<19:22> LBL: h12 ARG0: x8 ] in

In earlier days, we did not assume that every EP provides a unique distinguished variable as its ARG0, so the MRS for a sentence with repeated modifiers (often for intensifying or dramatic effect) would indeed be a bag, such as for the following:
“They live in a big, big house.” with RELS {… _big_a(x), _big_a(x), _house_n(x) …}
“He walked slowly, slowly into the room.” RELS {… _walk_v(e,x) _slow_a(e) _slow_a(e) …}
But MRSs produced by the ERG now aim to supply a unique ARG0 for every EP (apart from quantifiers), so at least ERS (English Resource Semantic) RELS lists seem to me now to be sets, not just bags.

Hi @ebender , my point was precisely to distinguish between the predicate symbol and the predication that uses one particular predicate symbol. We have a set of predications (each predication has a predicate symbol associated). If we take the predicate symbols for all predications in an MRS, than this is a multiset (or bag). In Pydelphin terms:

[p for p in sol.mrs.rels] # set
[p.predicate for p in sol.mrs.rels] # multiset