NEW SECRET MATH BENCHMARK STUMPS AI MODELS AND PHDS ALIKE

Summary from the TLDR newsletter

FrontierMath is a new mathematics benchmark that contains hundreds of expert-level problems which leading AI models are able to solve less than 2% of the time.