Question 25
Domain 3: Applications of Foundation ModelsWhich technique involves training AI models on labeled datasets to adapt the models to specific industry terminology and requirements?
Correct answer: B
Explanation
Fine-tuning means “updating the base FM weights on a smaller, task-specific or domain-specific dataset.” The source also says it “adapts the model's behavior without repeating the full pre-training cost,” which is why labeled industry data can tailor terminology and requirements.
Why each option is right or wrong
A. Data augmentation
Data augmentation creates modified training examples; it does not update model weights.
B. Fine-tuning
The question describes adapting a model using a labeled dataset for a narrow business domain, which matches the FM lifecycle’s fine-tuning stage: updating base model weights on a smaller, task-specific or domain-specific dataset. Under the source’s definition, this is the post-pretraining step used to specialize a foundation model for industry terminology and requirements, unlike pre-training which uses broad unlabeled corpora and massive compute.
C. Model quantization
Model quantization compresses model weights for efficiency, not domain adaptation.
D. Continued Pre-training
Continued Pre-training broadens general representations on large corpora before task-specific tuning.