Short Answer

Why a tighter human reference helps when error is already low

Question: A model has improved to about 8% error. Why would it still be useful to define a more careful human benchmark at 2% error instead of using a looser reference?

Sample answer: A tighter human benchmark gives a more informative target when the model is already fairly strong, so it helps the team see more clearly how much room for improvement remains and how to keep making progress.

Key points:

  • A precise human-level reference matters more when system error is already low.
  • It gives better guidance for the next round of improvements.

Rubric: The answer must explain that a more precise human benchmark provides better guidance or a better target for continuing to improve an already strong system.

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