Thanks these are helpful. Seems like "embedding bag" is used in ML libraries but not always in papers.

Edit: from a quick look neither of these is actually just an embedding bag, rather different approaches to incorporating subword information.