Short Answer

Why Recheck Apparently Correct Labels

Question: When cleaning labels in a validation set, why is it worth inspecting examples your classifier already marked correctly?

Sample answer: A case can look correct even when the recorded label and the model's prediction are both wrong in the same way. In that situation, the example is not actually evidence of a good label.

Key points:

  • A stored label may be wrong.
  • The model's prediction may also be wrong.
  • If both are wrong in the same direction, the example appears correct.

Rubric: The response should state that a seemingly correct example may hide the fact that both the label and the model prediction are incorrect.

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