Zero-shot-aware Training vs Fine-tuning for Adversarial Robustness in Multilingual Models

Large-scale pre-trained vision-language models like CLIP have demonstrated impressive performance across various tasks, and exhibit remarkable zero-shot generalization capability, while they are also vulnerable to imperceptible adversarial examples. Existing works typically employ adversarial training (fine-tuning) as a defense method against adversarial examples. However, direct application to the CLIP model may result in overfitting, compromising the model's capacity for generalization. In this paper, we propose Pre-trained Model Guided Adversarial Fine-Tuning (PMG-AFT) method, which leverag

Research goal: How do zero-shot-aware training techniques compare to standard fine-tuning in improving robustness against adversarial examples in multilingual models, as evaluated by accuracy on ANLI and adversarial QQP?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.5/10.