How It Works
Robustness test
A check of whether an AI model still behaves well when its input is changed a little: typos, rewording, noise, or inputs designed to trick it. A model that passes a benchmark but fails when the question is reworded isn't robust.
Origin · no single documented coiner
“Robustness” is an old term from statistics and engineering. Tests of AI models spread after research on “adversarial examples” (Szegedy et al., 2013) showed that tiny changes could fool image classifiers. No single coiner.