Skip to content

Latest commit

 

History

History
36 lines (19 loc) · 791 Bytes

section_rag.md

File metadata and controls

36 lines (19 loc) · 791 Bytes

Retrieval Augmented Generation (RAG)


Why?

ChatGPT hallucination

LLMs hallucinate

LLM is "lossy compression of its training data"

Updating LLMs is expensive


Fetch relevant documents from an external source

Provide relevant documents in the LLM prompt

Builds on LLM text generation capabilities

Increases accuracy

External source is easier to update


­ Retrieval Augmented Generation

(External source can be any information retrieval system)

Source: towardsdatascience.com