DeepYardDeepYard
R

RAG Retrieval Grading Prompt

Prompt for evaluating whether retrieved documents answer the user's question

Open SourceFree

About

A specialized prompt used in Corrective RAG (CRAG) pipelines to grade the relevance of retrieved documents. The prompt instructs the model to evaluate each document against the query, assign a relevance score, and flag when retrieval quality is too low — triggering fallback to web search. Critical for building self-correcting RAG systems that avoid hallucinating over irrelevant context.

Details

Use Caserag-evaluation
Target Modelsclaude, gpt-4, gemini

Tags

promptragevaluationgrading