What is statistical significance testing in model comparison?

It determines whether observed performance differences are due to chance.

Techniques like t-tests or McNemar’s test are used to compare models. The null hypothesis assumes no difference between models. A low p-value indicates statistically significant improvement.