Question: In one to three sentences, name one benefit of splitting a difficult prediction problem into separate models instead of training one model to produce the final label directly.

Sample answer: For a help-desk system, one model might first detect whether a message is about billing, and another could determine the specific billing issue. Because each model handles a narrower job, the pieces are usually easier to train and often need less labeled data than a single model that must learn everything at once.

Key points:

• Smaller subproblems are simpler to learn
• Each stage often needs fewer training examples than one all-in-one model

Rubric: Full credit for explaining that dividing the task into parts makes the individual models easier to train and reduces the amount of data needed. Partial credit for an answer that mentions only easier training or only lower data requirements.