#loss-vs-metric

Question 1: Single choice

A delivery-time model is trained by minimizing mean squared error, but the team reports its mean absolute error in minutes. Which statement is correct?

The loss guides the parameter updates, while the metric judges usefulness; the two may differ1
This is a mistake, because a model must always be evaluated with exactly the loss it was trained on2
The reported MAE is the quantity whose gradient was used to update the weights3
MAE cannot be computed for a model whose training loss was squared error4

Timed test