Data for RAG vs Training

1. What is the primary purpose of a RAG corpus in AI?

To provide authoritative and trustworthy sources for retrieval To train the model by providing labeled examples To supersede previous training datasets To generate synthetic data for modeling

2. Which of the following is a key requirement for training dataset preparation?

Maintaining class balance Ensuring every document is authoritative Minimizing duplicates Using only recent documents

3. What is a significant risk when using synthetic data for training?

It requires less governance than real data It can introduce biases from the generator model It is always better than naturally occurring data It guarantees a better performance in real-world scenarios

4. Why is having a held-out evaluation set critical before training a model?

To ensure diversity in the training dataset To prevent leaking test data into training To maximize the amount of labeled data available To validate the quality of synthetic data used

Continue: Beyond Text — Multimodal AI → Review this lesson Retake quiz