Research map / Adversarial robustness and security / Certified and empirical robustness
Contrastive Spectral Rectification: Test-Time Defense towards Zero-shot Adversarial Robustness of CLIP
Sen Nie, Jie Zhang, Zhuo Wang, Shiguang Shan, Xilin CHEN
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 Repulsion loss to Attraction loss (ablation) | CSR · General datasets | robust accuracy (%): 51.6 -> 69.8 | Improved |
| add Attraction loss to Repulsion loss (ablation) | CSR · General datasets | robust accuracy (%): 44.7 -> 69.8 | Improved |
| add Repulsion loss to Attraction loss (ablation) | CSR · Domain datasets | robust accuracy (%): 23.3 -> 40.3 | Improved |
| add Greedy Selection (ablation) | CSR · General datasets | robust accuracy (%): 68.5 -> 69.8 | Improved |
| add Greedy Selection (ablation) | CSR · Domain datasets | robust accuracy (%): 39.9 -> 40.3 | Improved |
| add Greedy Selection (ablation) | CSR · General datasets | clean accuracy (%): 78.6 -> 78.7 | No clear effect |
| use CSR test-time defense instead of TTC | CLIP ViT-B/16 · 16 datasets avg | robust accuracy (%): 24.5 -> 58.1 | Improved |
| use CSR instead of adversarial fine-tuning | CLIP ViT-B/16 · 16 datasets avg | clean accuracy (%): 43.6 -> 62.9 | Improved |
| use CSR defense on segmentation | CLIP zero-shot segmentation · VOC2010 under PGD | mIoU: 20.8 -> 37.5 | Improved |
| use CSR defense on captioning | LLaVA · MS-COCO under V-Attack | robust accuracy (%): 21.3 -> 64.3 | Improved |
| use CSR defense on VQA | LLaVA · VQA under V-Attack | robust accuracy (%): 22.6 -> 65.6 | Improved |
| use TTC defense on captioning and VQA | LLaVA · captioning and VQA | robust accuracy (%): 21.3 -> 24.1 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- R-TPT — R-TPT: Improving Adversarial Robustness of Vision-Language Models through Test-Time Prompt Tuning (CVPR 2025) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- CLIP is Strong Enough to Fight Back: Test-time Counterattacks towards Zero-shot Adversarial Robustness of CLIP (CVPR 2025 · similarity 0.89)
- TAPT: Test-Time Adversarial Prompt Tuning for Robust Inference in Vision-Language Models (CVPR 2025 · similarity 0.88)
- On the Zero-shot Adversarial Robustness of Vision-Language Models: A Truly Zero-shot and Training-free Approach (CVPR 2025 · similarity 0.88)
- CLIPTTA: Robust Contrastive Vision-Language Test-Time Adaptation (NeurIPS 2025 · similarity 0.88)
- R-TPT: Improving Adversarial Robustness of Vision-Language Models through Test-Time Prompt Tuning (CVPR 2025 · similarity 0.87)
- One Prompt Word is Enough to Boost Adversarial Robustness for Pre-trained Vision-Language Models (CVPR 2024 · 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.
