LearnAI

Navigation

Introduction to AI

Machine Learning

Deep Learning

NLP & LLMs

Computer Vision

Generative AI

Reinforcement Learning

AI Ethics & Governance

Tools & Frameworks

General

Prompt Patterns for RAG

Retrieval-Augmented Generation (RAG) combines a retrieval step with an LLM to ground outputs in external documents. Good prompt patterns make RAG systems more reliable and easier to evaluate.

Core patterns

  • Context + Instruction: Include the retrieved passages first, then a clear instruction: “Given the context below, answer concisely and cite the source IDs used.”
  • Concise Fusion: When many passages are returned, ask the model to synthesize only the most relevant 2–3 points.
  • Fallback Behavior: Explicitly instruct what to do when the context doesn’t contain an answer: return “No answer in provided context.”.

Formatting and citation

  • Request inline citations like (doc-123) and a short sources list at the end.
  • Limit token use by asking for summaries rather than verbatim repeats of context.

Evaluation tips

  • Test with out-of-context questions to confirm the model refuses to hallucinate.
  • Measure faithfulness by comparing generated citations to the actual retrieved passages.

Practical adoption of these patterns reduces hallucinations and improves traceability for RAG applications.