44. Evaluating generative models

Measure quality with task-specific benchmarks, human review, pairwise comparison, model-based judges, and modality-specific metrics. Design evaluation sets that resist contamination and reveal regressions across quality, cost, latency, and safety.