- Retrieval-Augmented Generation:
- Fine Tuning:
- Pros:
- Higher quality answers with respect to prompt engineering.
- Smaller input size during inference, since we don’t need to include the context and additional instructions in the prompt.
- Cons:
- It can be expensive
- The number of parameters of the model may not be sufficient to capture the entire knowledge we want to teach to it.
- Fine tuning is not additive, meaning it may be replacing some already existing knowledge with new information.
- Prompt Engineering: