FWIW I think a more generous view might be that there are different kinds of math, and most progress seems to be on “Olympiad-style” stuff.
2 1 53
will depue
@willdepue
Replying to @littmath
very interested how you divide the kinds of math and how the skills needed to succeed differ?
just guessing:
id naturally assume due to RL that math tasks fall on spectrum of verifiability and that has & will directly translate to models’ relative performance on such
i’d also
just guessing:
id naturally assume due to RL that math tasks fall on spectrum of verifiability and that has & will directly translate to models’ relative performance on such
i’d also
1 19