Short Answer

Why a coin flip is a bad debugging strategy for inference failures

Question: According to the inference-debugging framework, why is it a mistake to choose at random between fixing the retriever and fixing the ranking model when a pipeline fails?

Sample answer: Because those two failure modes point to different subsystems and therefore call for different fixes. If you work on the wrong subsystem, you may improve nothing and spend time without addressing the real cause of the bad output.

Key points:

  • Retriever errors and ranking errors require different interventions.
  • Random choice can send effort to the wrong place.
  • Working on the wrong part does not repair the failure.

Rubric: The answer should say that the failure type determines the needed action, and random selection can waste effort on the wrong component instead of correcting the cause.

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