Research map / Adversarial robustness and security / Attacks on text and graphs
Few-Shot Adversarial Prompt Learning on Vision-Language Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a few-shot adversarial prompt learning framework for vision-language models that learns adversarial text supervision end-to-end from adversarial examples, matching zero-shot adversarial robustness with 1% of training data.
Problem
Zero-shot adversarial robustness of vision-language models (e.g., CLIP) is limited by heavy adaptation cost, suboptimal handcrafted text supervision, and uncontrolled loss of natural generalization.
Approach
Adapt input sequences (visual and text prompts) with few-shot data, learning adversarially correlated text supervision end-to-end from adversarial examples, plus a training objective that makes multi-modal features consistent while making uni-modal features differentiated between natural and adversarial examples.
Key results
- Matches state-of-the-art zero-shot adversarial robustness with only 1% of training data, per the abstract.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use few-shot adversarial prompt learning with adversarially learned text supervision | Proposed framework on vision-language models · Adversarial robustness | adversarial robustness | Helped |
Experimental setup
Metrics: adversarial robustness accuracy zero-shot natural accuracy
Baselines: CLIP TeCoA PMG-AFT FT CoOp
Builds on: CLIP TeCoA PMG-AFT CoOp
Closest papers by meaning
- One Prompt Word is Enough to Boost Adversarial Robustness for Pre-trained Vision-Language Models (CVPR 2024 · similarity 0.91)
- Text-Guided Attention is All You Need for Zero-Shot Robustness in Vision-Language Models (NeurIPS 2024 · similarity 0.90)
- On the Zero-shot Adversarial Robustness of Vision-Language Models: A Truly Zero-shot and Training-free Approach (CVPR 2025 · similarity 0.89)
- Improving Zero-Shot Adversarial Robustness in Vision-Language Models by Closed-form Alignment of Adversarial Path Simplices (ICML 2025 · similarity 0.89)
- Language-Driven Anchors for Zero-Shot Adversarial Robustness (CVPR 2024 · similarity 0.89)
- R-TPT: Improving Adversarial Robustness of Vision-Language Models through Test-Time Prompt Tuning (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.