Research map / Adversarial robustness and security / Attacks on text and graphs
When CLIP Sees More, It Fights Back Harder: Multi-View Guided Adaptive Counterattacks for Test-Time Adversarial Robustness
Sunoh Kim, Daeho Um
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 affine transform to augmentations (ablation) | MAC · ten fine-grained avg | robust accuracy: 6.0 -> 45.2 | Improved |
| add Gaussian blur to augmentations (ablation) | MAC · ten fine-grained avg | robust accuracy: 42.3 -> 45.2 | Improved |
| add Gaussian noise to augmentations (ablation) | MAC · ten fine-grained avg | robust accuracy: 44.7 -> 45.2 | Improved |
| add color jitter to augmentations (ablation) | MAC · ten fine-grained avg | robust accuracy: 44.9 -> 45.2 | Improved |
| add Gaussian blur to augmentations (ablation) | MAC · ten fine-grained avg | clean accuracy: 58.8 -> 58.7 | No clear effect |
| use corruption degree instead of noise-driven deviation for weighting (ablation) | MAC · ten fine-grained avg | robust accuracy: +37.7% | Improved |
| use multiple views instead of single view (ablation) | MAC · ten fine-grained avg | robust accuracy: nearly 5 percentage points | Improved |
| add more than two views (ablation) | MAC · ten fine-grained avg | robust accuracy | No clear effect |
| increase counterattack iterations beyond K=4 (ablation) | MAC · ten fine-grained avg | robust accuracy | No clear effect |
| raise counterattack budget from 4 to 8 (ablation) | MAC · ten fine-grained avg, ViT-B/32 | robust accuracy: 33.7 -> 45.2 | Improved |
| raise counterattack budget from 4 to 8 (ablation) | TTC · ten fine-grained avg, ViT-B/32 | robust accuracy: 6.8 -> 3.9 | Hurt |
| scale backbone from ViT-B/32 to ViT-L/14 | MAC · fine-grained avg | robust accuracy: 45.2 -> 61.4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 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 4 of 5 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 0.91)
- TAPT: Test-Time Adversarial Prompt Tuning for Robust Inference in Vision-Language Models (CVPR 2025 · similarity 0.88)
- Robust CLIP: Unsupervised Adversarial Fine-Tuning of Vision Embeddings for Robust Large Vision-Language Models (ICML 2024 · similarity 0.87)
- R-TPT: Improving Adversarial Robustness of Vision-Language Models through Test-Time Prompt Tuning (CVPR 2025 · similarity 0.87)
- Mint: A Simple Test-Time Adaptation of Vision-Language Models against Common Corruptions (NeurIPS 2025 · similarity 0.86)
- Enhancing CLIP Robustness via Cross-Modality Alignment (NeurIPS 2025 · similarity 0.86)
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.
