Model Evaluation
Domain: Evaluation
🎯 Motivation
How do you know if an AI model is good? Evaluation measures performance on tasks - accuracy, speed, safety, fairness - using standardized methods.
📋 Concrete Example
Evaluate a model: run 1000 test questions. It answers 920 correctly. Accuracy: 92%. But also check: is it fast? Is it fair? Is it safe? Evaluation covers all dimensions.
🔗 Analogy
A teacher grading students - they test knowledge (accuracy), check handwriting speed (speed), ensure fair grading (fairness), and verify safety (no harmful behavior). Model evaluation grades AI models similarly.