Evaluation
LLM Evaluation
Measure retrieval, generation and agent quality. Catch regressions before users do.
Evaluation (eval) is the discipline of measuring LLM application quality systematically. This hub covers golden sets, retrieval metrics (recall@k, precision), generation metrics (faithfulness, answer relevance), LLM-as-judge, agent trajectory eval, CI/CD regression gates and online eval for drift detection.
Start Here
Begin with the fundamentals
The recommended first lesson for this topic.
Fundamentals
Tutorials
Architecture Patterns
Challenges
Key Concepts
Technologies