4 posts
AIME 2025 is the high school math competition that frontier AI models now use as a contamination-resistant reasoning benchmark. Here's how to read the scores.
GPQA is a graduate-level science benchmark designed to be unsolvable by Google search alone. Here's what the score actually means and how to use it.
Humanity's Last Exam is a 3,000-question benchmark designed to outlast frontier AI models. Here's what HLE actually tests and how to read the score.
SWE-Bench Verified is the benchmark that grades AI coding agents on real GitHub issues. Here's what it tests, what it misses, and how to read the scores.