RAG (Retrieval-Augmented Generation)

1. what is RAG?

    (1) RAG, which is the process that referencing reliable data from outside of training datasource

    (2) Optimizing LLM outputs before creating response

    (3) Unnecessary retraining data

    (4) Using Powerful LLM, Specific domain or internal data of the organization

 

2. Importance of the RAG

    (1) Problems of the LLM

       - Cannot predict about  LLM response

       - LLM trained data is static and trainied knowledge only valid on certain period of the time

       - Provide false information, if there are no specific answer

       - When users need specific answer or latest information, LLM response with outdated information or provides                      general informations

       - Inaccurate responses due to terminological confusion when using the same term to describe different things

 

     (2) Advantage of the RAG

        - Cost Effective

        - Latest informations

        - Strengthen user trust

        - Developer controllable 

3.  How it works?

    (1) External data creation

    (2) Searching relation information 

    (3) Extand of LLM prompt

    (4) External data update

 

 

Reference : https://aws.amazon.com/ko/what-is/retrieval-augmented-generation/