Context Relevance
For RAG systems, whether the retrieved documents actually contain information needed to answer the question.
Think of it like
Like bringing relevant books to a study session vs. random books from the library.
Example
Question: 'How do diffusion models work?' Retrieved doc: 'The weather in Paris today…' Context relevance: low.
How it actually works
If you retrieve irrelevant documents, the model has no chance to answer well. Context relevance measures the quality of retrieval. You can compute it automatically (are reference docs in the retrieved set?) or with LLM judges (do the docs contain relevant info?). Often the bottleneck in RAG.
For product teams
Context relevance determines whether RAG will work. Bad retrieval = bad answers, regardless of model quality.
For engineers
Measure retrieval precision and recall; optimize embedding models and retrieval strategies.
Related
- Answer Relevance — What improves by this.
- RAGAS — What this is part of.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome