ModelRefs / RAG Grounded Answer — Prompt Pattern
RAG Grounded Answer — Prompt Pattern
Answer strictly from retrieved context and cite sources inline.
Overview
Inject top-K retrieved chunks, instruct the model to answer only from them, refuse if unsupported, and cite by chunk ID.
When to use it: You need factual answers with provenance over a known corpus.
Pattern details
| Pattern class | retrieval |
|---|---|
| Difficulty | intermediate |
| Also known as | grounded prompt, rag answer prompt |
| Last reviewed | 2026-06-07 |
Known failure modes
- Ungrounded answer — Model answers from parametric knowledge. Mitigation: Add 'do not use outside knowledge' and an unsupported-answer eval.
- Over-refusal — Refuses when context is sufficient. Mitigation: Tune temperature and add positive examples.
When not to use it
- Stuffing 50+ chunks (overwhelms attention).
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to RAG Grounded Answer — Prompt Pattern.
Frequently asked questions
When should I use the RAG Grounded Answer prompt pattern?
You need factual answers with provenance over a known corpus.
What are common failure modes of RAG Grounded Answer?
Ungrounded answer • Over-refusal
Does RAG Grounded Answer work across models?
Yes when the anatomy is preserved — see model compatibility matrix on the pattern page.