Question: Explain why having substantial labeled data for intermediate stages can make a modular pipeline a practical design choice. In what way can this affect whether the pipeline performs better than a single direct model?

Sample answer: A modular pipeline becomes more attractive when each intermediate stage has enough labeled examples to learn its own task well. With that kind of supervision, the separate components can be trained carefully and can reach high accuracy on their specific subproblems. That can make the pipeline competitive with, or even better than, one model trained end to end, because the intermediate decisions are learned directly instead of being absorbed into one shared objective.

Key points:

• Enough labeled data exists for the intermediate stages
• Each component can be trained effectively on its own task
• The staged design can be competitive or superior because of targeted supervision

Rubric: The essay should explain that abundant data for the intermediate steps makes separate training feasible, and that this can make the modular approach potentially stronger than a single end-to-end model.