Unless their in-class performance increases as well, isn't that help "probably cheating"? (That's the "moral benchmark" I'd use, at least; if your collaboration resulted in you genuinely learning the material, it's probably not cheating.)