Question

A company deploys a customer-support chatbot built on a foundation model in Amazon Bedrock. Users report that the bot sometimes confidently states incorrect facts about the company's proprietary return policy, which was never part of the model's training data. The team wants to ground responses in the company's internal policy documents while minimizing these fabricated answers. Which approach best addresses this problem?

Accepted Answer

Implement Retrieval-Augmented Generation (RAG) so the model retrieves relevant policy documents and includes them in the prompt context before generating a response

Answer

Increase the temperature parameter so the model produces more varied and creative responses

Answer

Switch to a smaller foundation model that requires fewer tokens per request to reduce cost

Answer

Add more few-shot examples of polite greetings to the prompt to improve tone