Which concept improves accuracy by anchoring the model to verified data sources through retrieval?

Which concept improves accuracy by anchoring the model to verified data sources through retrieval?

• A. Shapley Values
• B. Surrogate Models
• C. Retrieval Augmented Generation (RAG)
• D. Utilitarianism

Answer: (C)

Retrieval-augmented generation focuses on grounding a language model’s outputs in external, verified data. By pairing a retriever that searches a corpus for relevant documents with a generator that conditions its response on those retrieved passages, the system can produce answers based on actual sources rather than relying solely on internal parameters. This helps reduce hallucinations and boosts factual accuracy, especially for current or domain-specific information, because the model uses retrieved documents to inform and support what it says. The other options don’t fit because Shapley Values are about attributing decisions to features, Surrogate Models are used to approximate complex models for interpretability, and Utilitarianism is an ethical theory, not a technical method for grounding AI responses.

Retrieval-augmented generation focuses on grounding a language model’s outputs in external, verified data. By pairing a retriever that searches a corpus for relevant documents with a generator that conditions its response on those retrieved passages, the system can produce answers based on actual sources rather than relying solely on internal parameters. This helps reduce hallucinations and boosts factual accuracy, especially for current or domain-specific information, because the model uses retrieved documents to inform and support what it says. The other options don’t fit because Shapley Values are about attributing decisions to features, Surrogate Models are used to approximate complex models for interpretability, and Utilitarianism is an ethical theory, not a technical method for grounding AI responses.