Retrieval Augmented Generation (RAG)
The technique combines the power of generative models with access to an external knowledge base. The model first retrieves relevant information from a database, then uses that information to generate an answer.
| Advantages | Disadvantages |
|---|---|
| Improves the accuracy and relevance of answers | Needs extra resources to store and process external data |
| Lets the model work with up-to-date information | Can be slower than purely generative approaches |
| Reduces the chance of model hallucinations | Complicates setup and integration with databases |
RAG prompt example:
Using the available database of recent scientific discoveries, describe the three most significant advances in quantum computing over the past year. For each advance, name the research group, the main results and the potential impact on technology.