Loading timeline…
20202020s
Retrieval-Augmented Generation (RAG)
Combining LLMs with external knowledge bases to prevent hallucinations.
Why It Was Important
Since an LLM's weights are static after training, RAG dynamically searches an external vector database (like Wikipedia or a corporate intranet) and forcefully injects that factual data into the prompt context window before the model answers. RAG became the absolute mandatory architecture for all enterprise AI deployments.
Who Invented It
Patrick Lewis et al. (Facebook AI)
Researchers bridging generative AI and deterministic search.
Applications
- Enterprise Chatbots
- Legal File Analysis
- Hallucination Prevention
Key Papers
- Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
Patrick Lewis · NeurIPS 2020
Videos
What is Retrieval-Augmented Generation (RAG)?
IBM Technology
What is Agentic RAG?
IBM Technology