SECAI Core · phase 5 of 15 · noun
Retrieval-augmented generation
A technique where a generative model retrieves relevant information from an external knowledge source and then uses that to generate its answer
The Explain card
- Plain English
- Retrieval-augmented generation (RAG) has a model fetch relevant information from an external knowledge source first, then use that material to generate its answer, instead of relying only on training memory.
- Example
- A company assistant answers "What is our laptop encryption policy?" by searching the internal policy repository, pulling the matching document, and composing a reply from it. The answer reflects current policy, not whatever the model absorbed during training.
- Why it matters
- RAG reduces hallucination and keeps answers current, but every retrieved document becomes model input. A poisoned page in the knowledge base can carry instructions or falsehoods straight into the answer.
- Hook
- RAG is open-book exam mode for the model, so guard the book.
Word knowledge
How the word is built, where it came from, and what it sits beside in memory.
In a sentence
The company chatbot runs retrieval-augmented generation against the policy vector index, then cites the returned encryption-policy chunk in its reply.
Why these words
- retrieval Old French retrouver, to find again names the hunt, so the producing hangs on a search rather than on memory
- augmented Latin augere, to increase, as a past participle marks that producing as boosted, so the hunt is extra apparatus rather than the whole job
- generation Latin generatio, a begetting, from generare, to beget supplies the head noun, the producing that the other two words qualify
Where it came from
- Origin
- English coinage by Patrick Lewis and colleagues, from Old French retrouver (to find again), Latin augere (to increase) as a past participle, and Latin generatio (a begetting)
- Entered the language
- 2020
- What changed
- Retrieval had meant finding again, then a library hunt, then a computer search. Augmented had meant increased, then boosted by extra apparatus, as in augmented reality. Generation had meant a begetting, then a cohort, then the producing of text. Lewis and colleagues hyphenated the first two onto generation in a 2020 paper title, and the three-word label passed from that title into everyday lab and product speech.
How it is spelled
- Pattern
- hyphen joins a noun to a past participle used as an adjective: retrieval-augmented
- Pattern
- retrieve drops its silent e before -al
- Pattern
- au spells /ɔː/ in augment, as it does in author and autumn
- Pattern
- -ation turns generate into a noun of activity
- Breaks the pattern
- retrieval-augmented keeps its hyphen as a unit; generation stays a free word
- Breaks the pattern
- the catalog spelling is three words; speech and papers often collapse it to RAG
Spelled like
- data-augmented
- survival
- author
- information
Broken into chunks
-
retriev
- retrieve
- retriever
-
al
- survival
- arrival
- withdrawal
-
aug
- augment
- August
- augmentation
-
ment
- fragment
- document
- monument
-
gen
- generate
- genesis
- progenitor
-
ation
- information
- creation
- operation
What it sits beside
Same subject
Same shape
- retrieval-augmented language model
- knowledge-augmented generation
- data-augmented training
Grounding family
Hyphenated -augmented compounds
- data-augmented
- knowledge-augmented
- context-augmented
Generation family
- generative AI
- text generation
- generative adversarial network
Where it sits in the deck
Phase 5: Interacting with Models: Prompting, APIs, and Retrieval
With a working model and clean data, learn the interaction layer — how users and systems communicate instructions and retrieve grounded answers.