Microsoft Certified: Azure AI Engineer AssociateImplement natural language processing solutionsHard
A large enterprise is implementing an internal knowledge base chatbot for its employees. The chatbot needs to answer a wide range of questions by retrieving information from a vast repository of internal documents, including PDFs, Word files, and wikis. It critically needs to provide source citations for its answers to build trust and allow employees to verify information. Which generative AI pattern is most suitable?
- AEncoder-Decoder Model
- BRetrieval Augmented Generation (RAG)
- CAbstractive Summarization
- DFine-tuned Generative Pre-trained Transformer (GPT)
Show answer & explanationAnswer & explanation
Correct answer: B. Retrieval Augmented Generation (RAG)
Retrieval Augmented Generation (RAG) is specifically designed for scenarios where answers must be grounded in a specific knowledge base and verifiable. By retrieving relevant documents first and then using an LLM to generate an answer based on those documents, RAG inherently supports providing source citations. This directly addresses the need for 'source citations' and answers from a 'vast repository of internal documents'.
Why the other options are wrong
- A. Encoder-Decoder models are a general architecture for sequence-to-sequence tasks, not a specific pattern for grounded QA with citations.
- C. Abstractive Summarization condenses text, it does not answer questions from a knowledge base or provide citations.
- D. A fine-tuned GPT might generate coherent answers but is prone to 'hallucinations' and cannot reliably provide source citations from an external repository without a retrieval mechanism.
Retrieval Augmented Generation (RAG) with Citations
A generative AI pattern that retrieves relevant information from a knowledge base to ground LLM responses, enabling the inclusion of source citations for verifiability.
- Crucial for factual accuracy and reducing hallucinations.
- Allows LLMs to answer questions using up-to-date, proprietary data.
- Inherently supports providing references to source documents.
Memory trick: RAG helps the AI 'show its work' by citing sources, building trust.