Well, it seems really hard to design an experiment to know if animals have abstraction capabilities.
Its actually quite easy… you try to teach them to understand english. If they can, they have the ability, if they cant, they dont. Even humans can understand languages without being taught translations through exposure.
The key is either animals are more advanced than us at symbolic manipulation in which case they would understand english and we wouldnt under stand them, or vice versa, we are the smarter of the two and we would understand their language (and see the symbolic manipulations within it).
The fact that they cant hold a conversation and do math is in and cant be taught, is in and of itself proof.
That said some chimps may be just barely capable (like koko who clearly could perform simple speech)… but beyond chimps it would seem easy to disprove by demonstration.
I think they must have some; knowing if a random animal belongs to your species or not is in itself a form of abstraction.
Seeing an animal and feeling a kinship towards it or wanting to engage with it differently certainly shows some awareness as to their similarity, but doesnt show a symbolic understanding, they do not create symbols that represent those concepts, they just know the concepts.
Since animals have very limited language abilities, it is not surprising their math abilities are (nearly?) undetectable.
I would only loosely call what most animals do language.. They are more like vocal signaling than language.. unless we talk about koko or something.