Ethical Robustness

Adversarial Training

A training method where the model simultaneously learns to resist attacks and maintain its ethical principles. This approach enhances robustness by exposing the system to hostile scenarios during its learning.

← Wstecz