Question

A retail company operates a customer support chatbot built on a pretrained large language model. Support agents complain that the bot frequently gives outdated answers about return policies, which the company updates every few weeks in an internal knowledge base. Leadership wants the bot to reflect the latest policies without retraining the model each time a policy changes. Which approach best meets this requirement?

Accepted Answer

Implement retrieval-augmented generation (RAG) so the model retrieves current policy documents from the knowledge base at query time

Answer

Fine-tune the pretrained model on the current policy documents each time the knowledge base is updated

Answer

Increase the maximum token limit of the model so it can process longer policy documents

Answer

Apply more detailed prompt engineering by embedding all historical policies directly in the system prompt