What is Retrieval Augmented Generation (RAG)?

It is a method that provides access to external knowledge sources to construct complex and knowledge-intensive structures. Developed by Meta AI, this method can generate answers using new information without model training.

made by GPT-4o

The RAG method is a combination of Retrieval and Generative models, enabling more accurate and contextually relevant responses.

• Retrieval models: Designed to retrieve information from a specific text source or database.
• Generative models: Designed to create new content according to a specific context.

Thanks to this method, the retraining step can be skipped and the answer can be generated from the source shown. Let’s explain with an example.

For example, let’s ask for information about a mission performed by SpaceX . As can be seen, it does not know the return date of CREW-8.

GPT-4o used

When we refer to the information on SpaceX’s own page, it can generate answers according to current information. Here we were able to get the most up-to-date answer without any cost or time loss to train the model. We were even able to provide a reference.