RAG (Retrieval-Augmented Generation)

Definition

In a RAG pipeline, the user question is first converted into an embedding and searched against a vector index of document chunks. The closest passages are optionally re-ranked by a reranking model and then passed as context to the language model. The model generates an answer grounded in the retrieved material, ideally with citations and source references. RAG keeps LLMs current, domain-specific, and auditable without retraining the base model.

How Swiss Knowledge Hub uses this term

Swiss Knowledge Hub is built as a RAG platform: every answer includes the source chunks that were used, complete with page number and a direct jump into the originating document. This reduces hallucinations and makes compliance audits easier.

Related terms

• LLM (Large Language Model)
• Embedding
• Vector database (vector store)
• Chunking
• Reranking
• Fine-tuning vs. RAG

Sources

1. Wikipedia: Retrieval-augmented generation — https://en.wikipedia.org/wiki/Retrieval-augmented_generation
2. Lewis et al. (2020) — Retrieval-Augmented Generation for Knowledge-Intensive NLP — https://arxiv.org/abs/2005.11401