47/50 for the grand prize is definitely a hard challenge. The best team in Progress Prize 1 only scored 29/50.
Good point. Maybe the hosts are assuming mathematicians and other folks will benefit from LLMs or the o1 OpenAI model? Kind of like raising the bar based on the technology available to the participants.
The 47/50 bar for the Overall Progress Prize was actually set in the first competition, it’s just that no one got it. Interestingly, the purse for the prize has doubled from “at least $794,624” to “at least $1,589,248”.
Perhaps intended as an Easter egg, those numbers are equal to 97*2^13 and 97*2^14, respectively.
I assume 97 <- 47+50? (although I'm not sure where the 13 and 14 respectively come from)
And the problems are much harder now.
1.5M is kaggle competition is also high. And progress is non linear in the field of reasoning.