Towards Fine-Grained Robustness: Attention-Guided Test-Time Prompt Tuning for Vision-Language Models

2026-05-19

AI Summary

This work addresses the limited robustness of current vision-language models under adversarial attacks and the tendency of existing test-time adaptation methods to degrade fine-grained semantic information. To overcome these limitations, the authors propose Attention-guided Test-Time Prompt Tuning (A-TPT), which introduces a gradient-based attention mechanism during inference for the first time. This mechanism identifies semantically discriminative regions that remain reliable under adversarial perturbations and leverages them to guide spatially adaptive data augmentation and multi-view ensemble strategies. By preserving fine-grained semantics while dynamically adapting to input-specific perturbations, A-TPT significantly improves both accuracy and robustness on both adversarial and clean samples, outperforming state-of-the-art test-time adaptation approaches.

Abstract

Vision-Language Models (VLMs), such as CLIP, have achieved significant zero-shot performance on downstream tasks with various fine-tuning adaptation methods. However, recent studies have proven that adversarial attacks can significantly degrade the inference ability of VLMs, posing substantial risks to their practical applications. Prevalent test-time adaptation methods typically rely on multi-view augmentation to implement various fine-tuning strategies, which struggle to identify semantic information and are prone to destroying discriminative regions in fine-grained scenarios. To address these limitations, we propose Attention-Guided Test-Time Prompt Tuning (A-TPT), a semantics-preserving method designed for test-time adaptation. We first refine the gradient attention rollout mechanism to identify semantically meaningful regions surviving under adversarial attacks. Furthermore, we leverage them to guide the spatially varying augmentation intensities and multi-view ensemble for prompt tuning and inference. Extensive experiments demonstrate that A-TPT outperforms existing test-time adaptation methods on both adversarial and clean data. Codes are available at https://github.com/SEU-VIPGroup/A-TPT .

Problem

• Vision-Language Models
• Adversarial Robustness
• Test-Time Adaptation
• Fine-Grained Recognition
• Semantic Preservation

Innovation

• test-time adaptation
• vision-language models
• adversarial robustness
• attention mechanism
• prompt tuning