Have we tried increasing cognition by selective breeding. Get mice best at maze to breed 100 descendants and repeat it few times, with varying food supply and survival difficulties.

This gets you mice that are better at navigating mazes. The connection between that and general cognition or learning capacity is not as robust as you would hope. Just as likely they simply have better peripheral vision or something.

https://en.m.wikipedia.org/wiki/Tryon%27s_Rat_Experiment