Retrieval augmented generation: Which deployment technique augments a model's responses with information

RAG grounds responses in externally retrieved documents at inference time without modifying model weights.

The question

Preparing for AIGP? Take the free 5-min readiness quiz →

1. Retrieval augmented generation (RAG), which supplements the model's prompt with relevant documents fetched from an external knowledge base at inference time. ✓

   Correct because RAG retrieves external content at query time to ground responses without retraining the model.
2. Fine-tuning, which updates the model's internal weights by training it further on a curated, task-specific labeled dataset before it is deployed.

   Plausible but wrong because fine-tuning changes the model's weights rather than retrieving external data at inference time.
3. An agentic architecture, which lets the model plan and autonomously call external tools in sequence to complete multi-step tasks on a user's behalf.

   Plausible but wrong because agentic tool-use is about action and orchestration, not grounding answers in retrieved documents.
4. Edge deployment, which runs the model locally on a device to reduce latency and keep data on-premises rather than in the cloud.

   Wrong because edge deployment concerns where the model runs, not augmenting responses with retrieved knowledge.

How to remember it

RAG grounds responses in externally retrieved documents at inference time without modifying model weights.

More Understanding How to Govern AI Deployment and Use questions

• A smaller classic model: Which model choice best fits these constraints? →
• To systematically identify and evaluate the potential: What is the primary purpose of performing or reviewing →
• Terms addressing data usage rights: When reviewing a vendor agreement for a third-party AI system, which →
• All 424 Understanding How to Govern AI Deployment and Use questions →