This is an embarrassing overreach.
We don't even have models that can win the far easier AMC, let alone AIME, USAMO, and then IMO.
We don't even have models that can win the far easier AMC, let alone AIME, USAMO, and then IMO.
I don't want to speculate, but it's not inconceivable that this will be achieved by AI within the future lifetime of XTX
AMC = multiple choice test, open to all grade school students.
AIME = open response test, all answers are numerical, open to students who score high enough on AMC only.
USAMO = USA Math Olympiad. IMO-style proof problems. Open only to top N scorers on AMC and AIME.