Active, anytime-valid risk controlling prediction sets

Rigorously establishing the safety of black-box machine learning models concerning critical risk measures is important for providing guarantees about model behavior.