NCA-GENL Question 159
Single answerYou have been tasked with building a chatbot for a customer support team using a large language model (LLM). The chatbot needs to provide accurate answers to customer queries based on a constantly updated knowledge base of product documentation, FAQs, and user guides. Which approach would be most suitable for implementing this use case?
- A
Fine-tune the LLM on the entire knowledge base to ensure it can answer customer queries accurately.
- B
Implement a Retrieval-Augmented Generation (RAG) framework to allow the chatbot to retrieve relevant information from the knowledge base dynamically.
- C
Use the LLM in its pre-trained state without customization and rely on its general knowledge to answer customer queries.
- D
Manually encode the knowledge base into the chatbot's responses as hardcoded templates for improved accuracy.
Show answer and explanation
Correct answer: B
Explanation
The RAG framework is the most appropriate choice for building a chatbot that relies on a constantly updated knowledge base. It allows the LLM to retrieve relevant information dynamically while generating accurate and contextually appropriate responses. This approach avoids the limitations of fine-tuning, pre-trained models without customization, and hardcoded templates.
- A. Incorrect.
Fine-tuning the LLM on the entire knowledge base is not ideal in this scenario because it requires significant computational resources and retraining every time the knowledge base is updated. Additionally, it can result in outdated or static responses.
- B. Correct.
A Retrieval-Augmented Generation (RAG) framework is highly suitable as it combines the LLM's natural language generation capabilities with dynamic retrieval of up-to-date, relevant information from the knowledge base. This ensures both accuracy and adaptability to updated content.
- C. Incorrect.
Using the LLM in its pre-trained state without customization is unlikely to meet the requirements because the model may lack specific domain knowledge from the knowledge base, leading to inaccurate or incomplete responses.
- D. Incorrect.
Hardcoding the knowledge base into the chatbot's responses as templates is inefficient and does not leverage the LLM's natural language understanding or ability to generate flexible, conversational responses.