Associate Professor, University of Auckland
1 paper at NeurIPS 2025
We propose CoAPT, a collaborative adversarial prompt tuning method that significantly improves the robustness and generalization of vision-language models against adversarial perturbations.
