Research map / Label-efficient and robust learning / OOD and anomaly detection
The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization
Dan Hendrycks, Steven Basart, Norman Mu, Saurav Kadavath, Frank Wang, Evan Dorundo, Rahul Desai, Tyler Zhu, Samyak Parajuli, Mike Guo, Dawn Song, Jacob Steinhardt and 1 more
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 DeepAugment + AugMix augmentation (ablation) | ResNet-50 · ImageNet-R | top-1 error (%): 63.9 -> 53.2 | Improved |
| add DeepAugment augmentation (ablation) | ResNet-50 · ImageNet-R | top-1 error (%): 63.9 -> 57.8 | Improved |
| add ImageNet-21K pretraining (10x labeled data) (ablation) | ResNet-50 · ImageNet-R | top-1 error (%): 63.9 -> 62.8 | No clear effect |
| add CBAM self-attention (ablation) | ResNet-50 · ImageNet-R | top-1 error (%): 63.9 -> 63.2 | No clear effect |
| add l_inf adversarial training (ablation) | ResNet-50 · ImageNet-R | top-1 error (%): 63.9 -> 68.6 | Hurt |
| add speckle noise augmentation instead of diverse augmentation (ablation) | ResNet-50 · ImageNet-R | top-1 error (%): 63.9 -> 62.1 | No clear effect |
| scale model from ResNet-50 to ResNet-152 (ablation) | ResNet · ImageNet-R | top-1 error (%): 63.9 -> 58.7 | Improved |
| add WSL pretraining on 1B Instagram images (ablation) | ResNeXt-101 32x8d · ImageNet-R | top-1 error (%): 57.5 -> 24.2 | Improved |
| add DeepAugment + AugMix augmentation (ablation) | ResNet-50 · ImageNet-C | mCE: 76.7 -> 53.6 | Improved |
| add DeepAugment + AugMix augmentation (ablation) | ResNeXt-101 32x8d · ImageNet-C | mCE: 66.7 -> 44.5 | Improved |
| add DeepAugment + AugMix augmentation (ablation) | ResNet-50 · Real Blurry Images | top-1 error (%): 58.7 -> 51.7 | Improved |
| add data augmentation (Style Transfer, DeepAugment, AugMix) (ablation) | ResNet-50 · SVSF (France) | classification error (%): 56.7 -> 59.1 | No clear effect |
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.
- + AugMix — AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty (ICLR 2020) · better on 2 of 3 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
- Measuring Robustness to Natural Distribution Shifts in Image Classification (NeurIPS 2020 · similarity 0.92)
- Benchmarking Low-Shot Robustness to Natural Distribution Shifts (ICCV 2023 · similarity 0.89)
- A Flat Minima Perspective on Understanding Augmentations and Model Robustness (AAAI 2026 · similarity 0.89)
- Unexplored Faces of Robustness and Out-of-Distribution: Covariate Shifts in Environment and Sensor Domains (CVPR 2024 · similarity 0.89)
- Addressing Model Vulnerability to Distributional Shifts Over Image Transformation Sets (ICCV 2019 · similarity 0.88)
- LAION-C: An Out-of-Distribution Benchmark for Web-Scale Vision Models (ICML 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.
