pull down to refresh

Scary that just 1 month ago, after evaluating o1, the great Terrence Tao-
anticipated that the benchmark would "resist AIs for several years at least," noting that the problems require substantial domain expertise and that we currently lack sufficient relevant training data.