Self-Corrective RAG (CRAG)
Data & AI Pipelines diagram · made in Skemio

What it shows
- Evaluator model scores retrieved document relevance before response generation.
- Triggers automated web search fallbacks when local retrieval confidence is low.
- Rewrites queries to strip noise and isolate the core information need.
The takeaway
Don't blindly trust retrieval; score relevance first, rewrite the query, and fallback to web when needed.
- #CRAG
- #RAG
- #Hallucination
- #GenerativeAI
- #MachineLearning
- #SystemDesign





