Question 4
Domain 1: Generative AI with LLMs and PromptingWhat is the primary advantage of transfer learning in LLM development?
Correct answer: B
Explanation
Transfer learning starts from a pretrained model, so developers reuse learned language patterns instead of training from scratch. This “reduces training time and computational requirements” because only the new task-specific layers or fine-tuning need additional work.
Why each option is right or wrong
A. Eliminates need for labeled data
B. Reduces training time and computational requirements
In LLM development, transfer learning leverages a pretrained foundation model and then adapts it to a downstream task, so the heavy cost of learning general language representations has already been paid. Because the model is not trained from random initialization, the process typically requires far fewer training steps, less GPU/TPU time, and lower compute expense than full pretraining from scratch.
C. Guarantees better performance
D. Removes all biases from models