NCA-GENL Question 158
Single answerA company wants to build a customer support chatbot using an LLM. They would like the chatbot to answer customer queries based on both predefined FAQs and dynamically fetched information from their internal knowledge base. Which approach would be most suitable for this use case?
- A
Fine-tune the LLM on the company's FAQ dataset and deploy it as a chatbot.
- B
Use a Retrieval-Augmented Generation (RAG) approach to integrate the LLM with the internal knowledge base.
- C
Train a custom LLM from scratch using the company's internal data.
- D
Use a summarization model to condense the company's internal knowledge base and feed it to the LLM.
Show answer and explanation
Correct answer: B
Explanation
The RAG approach is most suitable for this use case because it combines the generative capabilities of an LLM with the ability to retrieve relevant, up-to-date information from an external knowledge base. This ensures that the chatbot can handle both predefined FAQ responses and dynamic, real-time queries based on the company's internal data.
- A. Incorrect.
Fine-tuning the LLM on the FAQ dataset will improve its ability to answer predefined FAQ-based questions, but it will not dynamically fetch or utilize the internal knowledge base for updated or complex queries.
- B. Correct.
A Retrieval-Augmented Generation (RAG) approach is ideal because it enables the LLM to retrieve relevant information from the internal knowledge base in real-time, ensuring accurate and dynamic responses to customer queries.
- C. Incorrect.
Training a custom LLM from scratch is not practical for this scenario as it is resource-intensive, time-consuming, and unnecessary when pre-trained LLMs can be adapted using methods like RAG.
- D. Incorrect.
Using a summarization model to condense the knowledge base and feed it to the LLM might simplify the knowledge base but would not allow dynamic querying, which is essential for the chatbot's functionality.