Research map / Adversarial robustness and security / Attacks on text and graphs
Stabilizing Cross-Modal Bidirectional Attribution: Few-Shot Adversarial Prompt Tuning for Robust Vision-Language Models
Jun Feng, Shuhong Wu, Hong Sun, Pengfei Zhang, Bocheng Ren, Shunli Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add natural cross-entropy loss to adversarial loss (ablation) | CBA-FAPT · Base-to-new (16-shot, avg 11 datasets) | New natural accuracy (%): 46.18 -> 51.74 | Improved |
| add natural cross-entropy loss to adversarial loss (ablation) | CBA-FAPT · Base-to-new (16-shot, avg 11 datasets) | Base robust accuracy (%): 30.69 -> 35.72 | Improved |
| add visual attribution alignment loss (ablation) | CBA-FAPT · Base-to-new (16-shot, avg 11 datasets) | New natural accuracy (%): 51.74 -> 53.42 | Improved |
| add visual attribution alignment loss (ablation) | CBA-FAPT · Base-to-new (16-shot, avg 11 datasets) | New robust accuracy (%): 23.38 -> 24.17 | Improved |
| add textual attribution alignment (full cross-modal alignment) (ablation) | CBA-FAPT · Base-to-new (16-shot, avg 11 datasets) | New robust accuracy (%): 24.17 -> 25.44 | Improved |
| add textual attribution alignment (full cross-modal alignment) (ablation) | CBA-FAPT · Base-to-new (16-shot, avg 11 datasets) | Base robust accuracy (%): 36.45 -> 38.22 | Improved |
| full CBA-FAPT vs FAP on new classes | CBA-FAPT · Base-to-new (16-shot, avg 11 datasets) | New natural accuracy (%): 49.58 -> 55.74 | Improved |
| full CBA-FAPT vs FAP on new classes | CBA-FAPT · Base-to-new (16-shot, avg 11 datasets) | New robust accuracy (%): 21.86 -> 25.44 | Improved |
| full CBA-FAPT vs FAP on base classes | CBA-FAPT · Base-to-new (16-shot, avg 11 datasets) | Base natural accuracy (%): 70.52 -> 70.74 | No clear effect |
| CBA-FAPT vs FAP, cross-dataset natural accuracy | CBA-FAPT · Cross-dataset (ImageNet 16-shot, avg 11 datasets) | Natural accuracy (%): 47.76 -> 49.67 | Improved |
| CBA-FAPT vs FAP, cross-dataset robust accuracy | CBA-FAPT · Cross-dataset (ImageNet 16-shot, avg 11 datasets) | Robust accuracy (%): 24.23 -> 24.69 | Improved |
| CBA-FAPT vs FAP on Caltech101 robust accuracy | CBA-FAPT · Cross-dataset (Caltech101) | Robust accuracy (%): 65.43 -> 63.70 | Hurt |
Closest papers by meaning
- R-TPT: Improving Adversarial Robustness of Vision-Language Models through Test-Time Prompt Tuning (CVPR 2025 · similarity 0.89)
- One Prompt Word is Enough to Boost Adversarial Robustness for Pre-trained Vision-Language Models (CVPR 2024 · similarity 0.89)
- TAPT: Test-Time Adversarial Prompt Tuning for Robust Inference in Vision-Language Models (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 0.88)
- Few-Shot Adversarial Prompt Learning on Vision-Language Models (NeurIPS 2024 · similarity 0.88)
- Robust CLIP: Unsupervised Adversarial Fine-Tuning of Vision Embeddings for Robust Large Vision-Language Models (ICML 2024 · similarity 0.88)
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.
