Your question is LLM Evaluation in Clinical Settings. 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 would you evaluate the performance of a large language model in a clinical setting at Intuitive Surgical?
Discuss how you would define success, construct representative and adversarial evaluation data, measure clinical correctness and hallucination, and validate safety before deployment. Include offline and online evaluation, human review, latency and cost monitoring, prompt injection testing, and criteria for escalation or refusal.
