RAG pipeline
i
RAG pipeline
Retrieval-Augmented Generation: the system finds relevant documents and gives them to the model as evidence before it answers.
Move a question through retrieval, context assembly, and grounded generation.
Embed queryRetrieveGenerate
The question becomes an embedding used for semantic search.
QueryWhen was the first transformer paper published?
96%Attention Is All You Need
64%BERT paper
51%GPT-2 report
Grounded answerWaiting for retrieved context…