Lewis et al. (Meta AI) • 2020

Abstract

The original RAG paper. Lewis et al. proposed combining pre-trained language models with a retrieval mechanism that accesses external knowledge at generation time. This architecture became the foundation of every enterprise 'chat with your data' application, enabling models to ground their outputs in specific, verifiable information sources.

Warum Es Wichtig Ist

• Defined the RAG architecture now used across enterprise AI
• Enabled grounded generation from external knowledge bases
• Directly relevant to modern AI product development