Your question is LLM Evaluation Metrics. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How do you approach LLM evaluation? Describe the metrics you use to measure accuracy, latency, and cost.
Explain how you would build an evaluation process for a production LLM feature. Cover offline and online evaluation, test-set construction, human review, LLM-as-judge calibration, hallucination measurement, and cost and latency monitoring.
Provide concrete metric definitions and describe how evaluation results would guide model, prompt, retrieval, or infrastructure changes.