What type of study is this?

This is a Quantitative Study study.

October 12, 2025Open Access

FedAPT: Federated Adversarial Prompt Tuning for Vision-Language Models

View Full Paper

Ask AI

Bookmark

Authors

Federated adversarial prompt tuning improves robustness against attacks in models, highlighting class-aware information disparity.

FedAPT improves adversarial robustness in vision-language models under non-IID conditions, showing significant advantages.
Experiments reveal that FedAPT outperforms existing FPT methods, improving misclassification rates across various datasets.
The class-aware prompt generator integrates global and local label information, enhancing model defense against adversarial attacks.
Cross-layer sharing of prompt generation further strengthens prompt coupling, boosting robustness and effectiveness in real-world tasks.

Zhai et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark