Short Answer

State the main risk of fixing only misclassified dev set labels.

Question: What is the main problem with fixing only the dev-set examples that a classifier got wrong?

Sample answer: It makes the dev-set estimate too optimistic. Because only the cases that hurt the model are being corrected, the measured performance becomes biased upward and no longer gives a fair estimate of true performance.

Key points:

  • It introduces evaluation bias.
  • The bias is optimistic, so reported performance looks better than it should.

Rubric: The answer should state that selectively correcting only misclassified dev examples creates an optimistic bias that inflates the measured performance.

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

Related