Your question is A/B Testing for Model Variants. 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 use A/B testing to compare two model variants?
Explain how you would define the experimental unit, select a primary metric, randomize traffic, formulate hypotheses, determine sample size, analyze the result, and make a rollout decision. Address guardrail metrics, statistical significance, practical significance, repeated looks at the data, and possible interference between users or vehicles.