Research map / Label-efficient and robust learning / Recognition and long-tail
Spurious Features Everywhere - Large-Scale Detection of Harmful Spurious Features in ImageNet
Yannic Neuhaus, Maximilian Augustin, Valentyn Boreiko, Matthias Hein
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 |
|---|---|---|---|
| apply SpuFix (project out spurious feature directions) to robust ResNet50 | Robust ResNet50 · Spurious ImageNet | mAUC: 0.630 -> 0.763 | Improved |
| apply SpuFix to robust ResNet50 | Robust ResNet50 · ImageNet val | accuracy: 57.4% -> 56.8% | Hurt |
| transfer SpuFix to VOLO-D5 trained on ImageNet1k | VOLO-D5 512 · Spurious ImageNet | mAUC: 0.882 -> 0.907 | Improved |
| transfer SpuFix to models fine-tuned from ImageNet21k or trained on large data | BEiT-L/16, EVA-G/14 CLIP · Spurious ImageNet | mAUC: 0.921 -> 0.927 (BEiT-L) | Improved |
| pre-train on ImageNet21k then fine-tune on 1k | ViT-B AugReg · Spurious ImageNet | mAUC: 0.850 -> 0.917 | Improved |
| train on larger datasets (JFT, Instagram, LAION, MIM, CLIP) | ImageNet classifiers · Spurious ImageNet | mAUC | No clear effect |
| fine-tune ImageNet21k ViT-B on ImageNet1k | ViT-B AugReg · Spurious ImageNet | mAUC: 0.931 -> 0.917 | No clear effect |
Closest papers by meaning
- Let Samples Speak: Mitigating Spurious Correlation by Exploiting the Clusterness of Samples (CVPR 2025 · similarity 0.85)
- Complexity Matters: Feature Learning in the Presence of Spurious Correlations (ICML 2024 · similarity 0.84)
- Severing Spurious Correlations with Data Pruning (ICLR 2025 · similarity 0.84)
- A Sober Look at the Robustness of CLIPs to Spurious Features (NeurIPS 2024 · similarity 0.83)
- How Spurious Features are Memorized: Precise Analysis for Random and NTK Features (ICML 2024 · similarity 0.82)
- Generative Classifiers Avoid Shortcut Solutions (ICLR 2025 · similarity 0.81)
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.
