Question

A pharmaceutical company's support team uses a general-purpose LLM to answer employee questions about the company's internal clinical trial protocols. The model frequently gives answers that sound authoritative but contradict the actual protocol documents, which are updated monthly. Leadership asks for the most effective way to make answers accurate and traceable to the source documents without retraining the model each month. What should the team recommend?

Accepted Answer

Implement Retrieval Augmented Generation (RAG) that retrieves the relevant protocol documents at query time and grounds the model's responses in them

Answer

Fine-tune the model on the current protocol documents so the answers are baked into the model weights

Answer

Switch to a larger foundation model with more parameters to improve the accuracy of its answers

Answer

Lower the model's temperature setting so it produces more consistent and factual responses