Research map / Label-efficient and robust learning / Recognition and long-tail
Group Robust Classification Without Any Group Information
Christos Tsirigotis, Joao Monteiro, Pau Rodriguez, David Vazquez, Aaron Courville
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 |
|---|---|---|---|
| Use uLA instead of the strongest listed baseline, LC | cMNIST · cMNIST 0.5% and 5.0% | group-balanced accuracy (%): 71.25 -> 75.13; 91.16 -> 92.79 | Improved |
| Use uLA instead of LC | cMNIST · cMNIST 1.0% and 2.0% | group-balanced accuracy (%): 82.25 -> 81.80; 86.21 -> 84.79 | Hurt |
| Use uLA instead of the strongest listed baseline | cCIFAR10 · cCIFAR10 0.5%, 1.0%, 2.0%, and 5.0% | group-balanced accuracy (%): 34.56 -> 34.39; 38.30 -> 62.49; 47.81 -> 63.88; 57.32 -> 74.49 | Mixed |
| Use uLA instead of GroupDRO | Waterbirds · Waterbirds | worst-group accuracy (%): 91.4 -> 86.1 | Hurt |
| Use uLA instead of GroupDRO | CelebA · CelebA | worst-group accuracy (%): 88.9 -> 86.5 | Hurt |
| Use uLA instead of the strongest listed baseline for each split | sMPI3D · sMPI3D C=2, C=3, C=4, and C=5 | group-balanced accuracy (%): 31.94 -> 59.58; 48.93 -> 80.53; 71.94 -> 91.11; 94.62 -> 98.05 | Improved |
| Replace MoCoV2+ pretraining with BYOL (ablation) | uLA · cCIFAR10 1.0% | group-balanced accuracy (%): 62.49 -> 59.73 | No clear effect |
| Replace MoCoV2+ pretraining with Barlow Twins (ablation) | uLA · cCIFAR10 1.0% | group-balanced accuracy (%): 62.49 -> 50.08 | Hurt |
| Use SSL-pretrained backbone rather than training from scratch (ablation) | uLA and the compared training paradigms · CelebA, cCIFAR10, and sMPI3D | robust accuracy | Improved |
| Apply logit adjustment rather than vanilla cross-entropy finetuning (ablation) | pretrained backbone · CelebA and cCIFAR10 | robust accuracy | Improved |
| Use uLA rather than supervised-bias sLA (ablation) | logit-adjustment training · CelebA and cCIFAR10 | robust accuracy | No clear effect |
| Use uLA rather than supervised-bias sLA (ablation) | logit-adjustment training · sMPI3D | robust accuracy | Mixed |
Closest papers by meaning
- The Group Robustness is in the Details: Revisiting Finetuning under Spurious Correlations (NeurIPS 2024 · similarity 0.88)
- ExMap: Leveraging Explainability Heatmaps for Unsupervised Group Robustness to Spurious Correlations (CVPR 2024 · similarity 0.88)
- Improving Subgroup Robustness via Data Selection (NeurIPS 2024 · similarity 0.88)
- Improving Group Robustness on Spurious Correlation Requires Preciser Group Inference (ICML 2024 · similarity 0.88)
- Improving Robustness to Multiple Spurious Correlations by Multi-Objective Optimization (ICML 2024 · similarity 0.87)
- Rank-Constrained Adaptation for Reliable Real-World Performance (NeurIPS 2026 · 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.
