Research map / Architectures and efficiency / Attention and transformers
Improving Robustness of Vision Transformers by Reducing Sensitivity To Patch Corruptions
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Apply RSPC to RVT-S | RVT-S · CIFAR-10-C | accuracy: 92.35 (+0.00) -> 94.14 (+1.79) | Helped |
| Apply RSPC to RVT-S | RVT-S · CIFAR-100-C | accuracy: 73.43 (+0.00) -> 74.94 (+1.51) | Helped |
| Apply RSPC to FAN-S-Hybrid | FAN-S-Hybrid · CIFAR-10-C | accuracy: 93.14 (+0.00) -> 94.59 (+1.45) | Helped |
| Apply RSPC to FAN-S-Hybrid | FAN-S-Hybrid · CIFAR-100-C | accuracy: 74.19 (+0.00) -> 75.72 (+1.53) | Helped |
| Apply RSPC to FAN-S-Hybrid | FAN-S-Hybrid · ImageNet-A | accuracy: 33.9 (+0.0) -> 36.8 (+2.9) | Helped |
| Apply RSPC to FAN-S-Hybrid | FAN-S-Hybrid · ImageNet-C | mCE: 48.5 (-0.0) -> 47.5 (-1.0) | Helped |
| Apply RSPC to FAN-S-Hybrid | FAN-S-Hybrid · ImageNet-P | mFR: 34.5 (-0.0) -> 33.5 (-1.0) | Helped |
| Apply RSPC to FAN-B-Hybrid | FAN-B-Hybrid · ImageNet-A | accuracy: 39.6 (+0.0) -> 41.1 (+1.5) | Helped |
| Apply RSPC to FAN-B-Hybrid | FAN-B-Hybrid · ImageNet-C | mCE: 46.1 (-0.0) -> 44.5 (-1.6) | Helped |
| Apply RSPC to FAN-B-Hybrid | FAN-B-Hybrid · ImageNet-P | mFR: 31.3 (-0.0) -> 30.0 (-1.2) | Helped |
| Choose corrupted patches with the learned corruption model rather than randomly (ablation) | RSPC on FAN-B-Hybrid · ImageNet | mCE: 45.6 (-0.5) -> 44.5 (-1.6) | Helped |
| Add clean-and-corrupted training examples instead of clean-only training (ablation) | FAN-B-Hybrid · ImageNet | mCE: 46.1 (-0.0) -> 45.8 (-0.3) | Helped |
Closest papers by meaning
- Understanding and Improving Robustness of Vision Transformers through Patch-based Negative Augmentation (NeurIPS 2022 · similarity 0.90)
- Certified Patch Robustness via Smoothed Vision Transformers (CVPR 2022 · similarity 0.90)
- Improving Interpretation Faithfulness for Vision Transformers (ICML 2024 · similarity 0.89)
- Towards Understanding and Improving Adversarial Robustness of Vision Transformers (CVPR 2024 · similarity 0.87)
- Transformers Learn Low Sensitivity Functions: Investigations and Implications (ICLR 2025 · similarity 0.86)
- Harnessing Edge Information for Improved Robustness in Vision Transformers (AAAI 2024 · 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.