Research map / Label-efficient and robust learning / Test-time adaptation
Improving robustness against common corruptions by covariate shift adaptation
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 |
|---|---|---|---|
| adapt BN statistics on test corruption samples | Vanilla ResNet-50 · ImageNet-C | mCE: 76.69 -> 62.24 | Helped |
| adapt BN statistics | DeepAug+AM ResNet-50 · ImageNet-C | mCE: 53.55 -> 45.36 | Helped |
| adapt BN statistics | DeepAug+AM+RNXt101 · ImageNet-C | mCE: 44.5 -> 38.0 | Helped |
| adapt BN statistics | ResNeXt101 with IG-3.5B weakly supervised pretraining · ImageNet-C | mCE: 51.70 -> 51.64; 45.7 -> 47.3 | Mixed |
| replace BN with GroupNorm or Fixup (no adaptation) (ablation) | ResNet-50 · ImageNet-C | mCE: 76.7 -> 72.00 | Helped |
| adapt BN statistics when corruptions are shuffled across batch (ablation) | ResNet-50 · Mixed IN-C | top-1 error: 61.08 -> 60.87 | No clear effect |
| adapt BN statistics | ResNet-50 · ObjectNet | top-1 error: 78.15 -> 75.96 | Helped |
| adapt BN statistics | robust models · ImageNet-R | top-1 error: 47.9 -> 44.0 | Helped |
Closest papers by meaning
- Suppress and Diversify: Refining Robust Pathways for Corruption Robustness (ICML 2026 · similarity 0.91)
- LAION-C: An Out-of-Distribution Benchmark for Web-Scale Vision Models (ICML 2025 · similarity 0.90)
- Adaptive Sensitivity Analysis for Robust Augmentation against Natural Corruptions in Image Segmentation (ICML 2025 · similarity 0.88)
- VLM-RobustBench: A Comprehensive Benchmark for Robustness of Vision-Language Models (ICML 2026 · similarity 0.87)
- ImageNet-D: Benchmarking Neural Network Robustness on Diffusion Synthetic Object (CVPR 2024 · similarity 0.87)
- Foundation Model-oriented Robustness: Robust Image Model Evaluation with Pretrained Models (ICLR 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.