As older researchers will often tell you: the job of a reviewer is to refuse your paper. It doesn't have to be completely black and white though. Maybe they did

As older researchers will often tell you: the job of a reviewer is to refuse your paper. It doesn't have to be completely black and white though. Maybe they did the maths and the results were middling/unclear, their calculations were filled with assumptions that they didn't want to/didn't have time to defend. In ML it's not uncommon to drop datasets that you evaluated on but that don't achieve SOTA because the reviewer will focus on that and ignore your other results. An article must have a clear concise point.