LLM Retrieval Augmented Generation
LLM Retrieval Augmented Generation — Grounding model answers in retrieved documents.
Definition
LLM Retrieval-Augmented Generation is RAG applied specifically to large language models: retrieve passages, stuff or rank them into context, then generate an answer that cites those passages.
Why it matters
It is the practical form of grounding used in Memory/Codex search. Quality depends on chunking, embedding recall, and whether the retrieved text actually contains a definition — not on model size alone.
Case
Wiki Ask RAG that only retrieves “Title — AI concept.” will paraphrase that stub. The failure is corpus thinness, not “the model is dumb.” Enrich the entry with Definition/Case, re-embed, then re-ask.
How to read it
Inspect top citations before trusting the prose. Prefer answers that start with a plain definition and admit missing evidence over long Alpha Factory narratives built on empty stubs.