Question

A retail company has built a generative AI assistant to answer customer questions about return policies. Before deploying it to production, the AI transformation leader wants to confirm the model produces accurate, relevant, and appropriate responses across a range of realistic customer queries. Which machine-learning lifecycle activity should the team perform to meet this goal?

Accepted Answer

Systematically evaluate the model's responses against a curated test set of representative queries with defined quality criteria before release

Answer

Increase the maximum output token limit so the assistant can generate longer, more detailed answers

Answer

Skip evaluation and rely on live customer feedback after deployment to identify any problems

Answer

Retrain the base foundation model from scratch using the company's return-policy documents