Pipeline
How every task is built, from professional sources to an accepted task. Click a step for what it does.
Leaderboard
Pass rate, cost and time per task for each model. A task passes only when all of its scoring criteria are met.
Task bank preview
A preview of finished tasks from the pipeline: the question as delivered, its reference analysis, scoring criteria and supporting quotations, each linked to the original source. The full task collection will be released later.