Short Answer

What information should a person use when checking one stage of a pipeline against human performance?

Question: When comparing one step in a processing pipeline to human-level performance, what limitation should be placed on the human reviewer?

Sample answer: The human should be allowed to see only the same inputs that reach that one step from earlier stages, not the full original data source. That keeps the comparison fair.

Key points:

  • The human uses the same input available to the component.
  • The human does not get access to the raw source data.
  • Earlier-stage outputs are the only permitted inputs.

Rubric: The answer must state that the human evaluator is restricted to the outputs of the earlier pipeline stages, so the comparison matches the information available to the component.

0

1

Updated 2026-08-12

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Machine Learning Yearning @ DeepLearning.AI