AI Glossary

Adversarial training

What does “Adversarial training” mean?

Adversarial training: Training with adversarial examples or attacks to improve robustness against them.

Adversarial training visual guide: Training with adversarial examples or attacks to improve robustness against them.
Visual guide Open full size (new tab)

What is an example of “Adversarial training”?

A vision model is trained on both ordinary images and carefully perturbed versions.

Keep learning

See also

AlignmentTermAI safetyTermAgent guardrailsTermAbstentionTermAdversarial exampleTerm

Same topic

Data exfiltrationTermIndirect prompt injectionTermAI biasTermAI regulationTermAblation studyTerm