NCA-GENL Question 21
Select 3You are tasked with building a retrieval-augmented generation (RAG) system to assist customer support agents in answering technical queries. Which of the following components are essential for implementing a functional RAG pipeline?
- A
A vector database to store and retrieve relevant documents
- B
A fine-tuned large language model (LLM) for generating responses
- C
A fixed rule-based system for generating answers
- D
A document retriever to search for relevant context from stored data
- E
A GPU-accelerated training pipeline for real-time model updates
Show answer and explanation
Correct answers: A, B, D
Explanation
A retrieval-augmented generation (RAG) pipeline combines document retrieval with language model generation to provide contextually relevant and accurate answers. Essential components include a vector database for storing document embeddings, a retriever to fetch relevant documents, and a fine-tuned LLM for generating responses. Optional components like GPU-based training pipelines are only needed for model updates, not for the core RAG functionality.
- A. Correct.
A vector database is critical for storing embeddings of documents and efficiently retrieving relevant context during query processing, making it an essential component of a RAG system.
- B. Correct.
A fine-tuned large language model (LLM) is required to generate coherent and contextually appropriate responses after retrieving relevant information, making it a key part of the pipeline.
- C. Incorrect.
A fixed rule-based system is not necessary for RAG, as the generation of responses is handled by the LLM, which dynamically generates answers based on context.
- D. Correct.
A document retriever is essential for identifying relevant documents from the vector database, enabling the LLM to use this context for generating accurate responses.
- E. Incorrect.
While a GPU-accelerated training pipeline can help with training new models or fine-tuning existing ones, it is not a mandatory component for deploying a RAG system, as pre-trained models can be used effectively.