Research map / Label-efficient and robust learning / Recognition and long-tail
Avoiding spurious correlations via logit correction
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 |
|---|---|---|---|
| Use logit correction instead of the strongest listed baseline | LC · Colored MNIST, 0.5% minority | group-balanced accuracy: 65.22 (4.41) -> 71.25 (3.17) | Helped |
| Use logit correction instead of Group DRO | Corrupted CIFAR-10 · 0.5% minority | group-balanced accuracy: 33.44 -> 34.56 (0.69) | Helped |
| Use logit correction instead of Group DRO | Corrupted CIFAR-10 · 1.0% minority | group-balanced accuracy: 38.30 -> 37.34 (1.26) | Hurt |
| Use logit correction instead of Group DRO | Corrupted CIFAR-10 · 2.0% minority | group-balanced accuracy: 45.81 -> 47.81 (2.00) | Helped |
| Use logit correction instead of Group DRO | Corrupted CIFAR-10 · 5.0% minority | group-balanced accuracy: 57.32 -> 54.55 (1.26) | Hurt |
| Add Group MixUp to ERM (ablation) | bFFHQ · bFFHQ | minority-group accuracy: 56.87 -> 63.22 | Helped |
| Add logit correction to ERM (ablation) | bFFHQ · bFFHQ | minority-group accuracy: 56.87 -> 66.51 | Helped |
| Add Group MixUp to logit correction (ablation) | bFFHQ · LC loss | minority-group accuracy: 66.51 -> 69.67 | Helped |
| Vary the GCE q parameter (ablation) | LC on Waterbirds · Waterbirds | worst-group accuracy: 84.2; 88.6; 88.9; 89.9; 90.7; 89.4 | Mixed |
| Use logit correction instead of reweighting | ImageNet-pretrained ResNet-18 linear classifier on Corrupted CIFAR-10 · 5% minority | group-balanced accuracy: 23.75% -> 30.20% | Helped |
| Use logit correction instead of Fisher-Consistent Reweighting | Model trained with ground-truth group labels · Ground-truth group labels | test accuracy: 42.13 -> 54.31 | Helped |
Closest papers by meaning
- Mitigating Spurious Correlations via Disagreement Probability (NeurIPS 2024 · similarity 0.87)
- Let Samples Speak: Mitigating Spurious Correlation by Exploiting the Clusterness of Samples (CVPR 2025 · similarity 0.87)
- Robust Logit Adjustment for Learning with Long-Tailed Noisy Data (AAAI 2025 · similarity 0.86)
- Improving Group Robustness on Spurious Correlation Requires Preciser Group Inference (ICML 2024 · similarity 0.85)
- Understanding Domain Generalization: A Noise Robustness Perspective (ICLR 2024 · similarity 0.85)
- Boosting Test Performance with Importance Sampling--a Subpopulation Perspective (AAAI 2025 · similarity 0.84)
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.