Precision

Precision is an evaluation metric measuring the share of a model's positive predictions that are actually correct, answering how many of the items it flagged were right.

Precision sits in the Evaluation & safety part of the Agentik {OS} glossary, which defines the words used to build and run AI agent systems.