It would have been nice to have a list with an example for each of the 23 tasks. Have I missed it?
They have a lot of examples in the appendix, just scroll to after the references.
On page 24, figure 6 right, I think the “gold answer” is incorrect. It should be 160. It may have been copy-paste-not edited from the gold answer on page 22, figure 4 right.
Seeing those insane advances really makes me itch to go back into academia. AI that can do math really feels like we're about to crack 'it'.