RAG (Retrieval-Augmented Generation)
An AI system architecture in which a model retrieves relevant documents or data from an external store at query time and includes them in its prompt, rather than having been trained directly on that data. Removing a document from the retrieval store generally prevents its future retrieval, a retention advantage over fine-tuning, though the retrieval store's embeddings, any cached prior responses that already included the document's content, backup copies of the index, and vendor training-on-query terms remain separate considerations.
Defined in 2 GAGE programs, which carry 9 distinct definitions of it. The wording above is taught in AI Data Governance: The Data Chair.
How each discipline defines it
The same term does different work depending on who is using it. These are the definitions as each program teaches them, unedited.
An AI system architecture in which a model retrieves relevant documents or data from an external store at query time and includes them in its prompt, rather than having been trained directly on that data. Removing a document from the retrieval store generally prevents its future retrieval, a retention advantage over fine-tuning, though the retrieval store's embeddings, any cached prior responses that already included the document's content, backup copies of the index, and vendor training-on-query terms remain separate considerations.
a technique in which a system searches an external document store or database at run time and incorporates relevant results into its response, without retraining the underlying model.
An architecture where a language model's response is grounded by first retrieving relevant text from a vector store (using embedding similarity search) and including that retrieved text as context for the model's generation. (see Topic 4.4)
An AI architecture where a system searches a corpus of documents for content relevant to a user's question, retrieves the most relevant results, and feeds that retrieved content to a language model as context to generate an answer.
An AI architecture where a model retrieves relevant documents from an external corpus at query time and uses them as context to generate a response, rather than relying solely on knowledge encoded during training.
Where it is taught
The exact lessons this term appears in. The first 7 topics of every program are free with a free account.
- The Eight Components I · The Agent, Deconstructed, Agentic AI Governance: Applied Mastery
- Retention versus the model that memorized: deleting data a model already learned from · Consent, Purpose, and the Law of Data, AI Data Governance: The Data Chair
- Embeddings are data too: the vector store as a personal-data system · Feeding the Machines, AI Data Governance: The Data Chair
- The injectable corpus: documents that attack the AI that reads them · Poison, Leaks, and the Adversary, AI Data Governance: The Data Chair
- Membership inference and extraction: what a model reveals about its training data · Poison, Leaks, and the Adversary, AI Data Governance: The Data Chair
- Red-teaming your own corpus: finding the poison before the attacker uses it · Poison, Leaks, and the Adversary, AI Data Governance: The Data Chair
- Least privilege in the estate: who can read what, and the analyst who can read everything · Access and the Keys, AI Data Governance: The Data Chair
- The taxonomy of data incidents: breach, leak, poison, drift, and wrongful training · Data Incidents, AI Data Governance: The Data Chair
- The deletion that was not: personal data found in a model trained last year · Data Incidents, AI Data Governance: The Data Chair
Terms it appears with
Not an alphabetical neighbourhood: these are the terms taught in the same lessons, ranked by how often they appear together.