Adversarial Vulnerabilities in Zero-Shot Learning: An Empirical Examination

This ArXiv article examines the robustness of zero-shot learning models against adversarial attacks, a critical area for ensuring model reliability and safety. The empirical study likely provides valuable insights into the vulnerabilities of these models and potential mitigation strategies.

• •Identifies vulnerabilities in zero-shot learning models under adversarial attacks.
• •Investigates the impact of attacks at both class and concept levels.
• •Likely offers insights for improving the robustness of zero-shot learning systems.