Standardized methods for evaluating AI system performance. Enables comparison across models, tasks, and metrics.