Question 27
Domain 3: Applications of Foundation ModelsAn AI practitioner needs to improve the accuracy of a natural language generation model. The model uses rapidly changing inventory data. Which technique will improve the model's accuracy?
Correct answer: C
Explanation
Retrieval Augmented Generation improves accuracy when inventory data changes quickly because the model can fetch the latest facts at query time instead of relying on stale training data. The source says GenAI accuracy can be improved through “retrieval,” and that embeddings plus a vector database enable “similarity retrieval” of current context before the model responds.
Why each option is right or wrong
A. Transfer learning
Transfer learning adapts a foundation model to downstream tasks through prompting, retrieval, or fine-tuning.
B. Federated learning
Federated learning trains across distributed data sources while keeping data local.
C. Retrieval Augmented Generation (RAG)
AWS’s GenAI guidance identifies retrieval as a way to improve output accuracy by supplying the model with current context at inference time, rather than depending on whatever was learned during pretraining. For rapidly changing inventory data, this matters because stale training data will lag behind the live stock state; using embeddings plus a vector database enables similarity retrieval of the latest records before generation, which is the mechanism described in the source material.
D. One-shot prompting
One-shot prompting uses a single example in the prompt; it does not fetch updated external facts.