NCA-GENL Question 22
Single answerA company wants to implement a customer support system using a generative AI model. The system should provide concise and accurate answers to customer queries using proprietary company documents as a knowledge base. Which approach is most appropriate for building this use case?
- A
Fine-tune a large language model (LLM) directly on the company's proprietary documents.
- B
Use a retrieval-augmented generation (RAG) pipeline to retrieve relevant documents and generate responses.
- C
Train a new language model from scratch using the company’s data.
- D
Directly prompt a pre-trained LLM with customer queries without additional customization.
Show answer and explanation
Correct answer: B
Explanation
The retrieval-augmented generation (RAG) approach is ideal for building a customer support system that uses proprietary documents. It enables retrieving relevant information from a knowledge base and generating accurate, concise responses using a pre-trained LLM. This approach balances efficiency, accuracy, and adaptability, making it the most suitable choice for this scenario.
- A. Incorrect.
Fine-tuning an LLM on proprietary documents can be resource-intensive, time-consuming, and may not generalize well to new or unseen queries. It is not the most efficient approach for this use case.
- B. Correct.
A retrieval-augmented generation (RAG) pipeline combines the generative capabilities of an LLM with the ability to retrieve relevant information from a specific knowledge base, making it highly suitable for scenarios involving proprietary data.
- C. Incorrect.
Training a new language model from scratch is impractical due to the significant computational resources, data, and expertise required. It also lacks the efficiency of leveraging pre-trained LLMs and retrieval methods.
- D. Incorrect.
Using a pre-trained LLM without customization might yield general responses, but it cannot accurately incorporate proprietary company data, which is a key requirement for this use case.