Source knowledge is external information supplied to a model through knowledge stores rather than learned within its parameters. In Retrieval-Augmented Generation (RAG), it complements parametric knowledge with current, verifiable evidence. section 1.2 section 1.0
RAG uses source knowledge to inform generated answers: query processing shapes retrieval, a knowledge-base search finds context, and generation synthesizes that context with the user’s question. This can provide freshness and source attribution, but retrieving the right document does not guarantee that the answer is grounded in it. section 1.2 section 1.0
A policy document within reach
Air Canada’s chatbot answered a bereavement-fare question incorrectly even though the correct policy was linked in the same message. The case illustrates that access to source knowledge is different from using it faithfully. section 1.0
Figure 1.1 The RAG workflow: from user queries to grounded responses through retrieval and generation.

Trace how a user query becomes a grounded response through query processing, retrieval, context assembly, and generation. section 1.2
