Research map / Label-efficient and robust learning / Recognition and long-tail
Simple and Fast Group Robustness by Automatic Feature Reweighting
Shikai Qiu, Andres Potapczynski, Pavel Izmailov, Andrew Gordon Wilson
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 AFR instead of ERM | ResNet-50 · Waterbirds | worst-group accuracy: 72.6 -> 90.4_{\pm1.1} | Improved |
| Use AFR instead of ERM | BERT · MultiNLI | worst-group accuracy: 67.9 -> 73.4_{\pm0.6} | Improved |
| Use AFR instead of ERM | DenseNet-121 · Chest X-Ray | worst-group accuracy: 4.4 -> 56.0_{\pm3.4} | Improved |
| Use AFR instead of ERM | ResNet-50 · CelebA | worst-group accuracy: 47.2 -> 82.0_{\pm0.5} | Improved |
| Use AFR instead of ERM | BERT · CivilComments | worst-group accuracy: 57.4 -> 68.7_{\pm 0.6} | Improved |
| Remove ℓ2 regularization from AFR (ablation) | AFR · Waterbirds | worst-group accuracy: 90.4_{\pm0.1} -> 90.1_{\pm0.7} | No clear effect |
| Remove ℓ2 regularization from AFR (ablation) | AFR · CelebA | worst-group accuracy: 82.0_{\pm0.5} -> 81.3_{\pm1.8} | No clear effect |
| Remove ℓ2 regularization from AFR (ablation) | AFR · MultiNLI | worst-group accuracy: 73.4_{\pm0.6} -> 71.1_{\pm1.4} | Hurt |
| Remove second-stage early stopping from AFR (ablation) | AFR · Waterbirds | worst-group accuracy: 90.4_{\pm0.1} -> 89.9_{\pm0.7} | No clear effect |
| Remove second-stage early stopping from AFR (ablation) | AFR · CelebA | worst-group accuracy: 82.0_{\pm0.5} -> 80.0_{\pm1.6} | Hurt |
| Remove second-stage early stopping from AFR (ablation) | AFR · MultiNLI | worst-group accuracy: 73.4_{\pm0.6} -> 71.1_{\pm1.4} | Hurt |
| Replace exponential weights with (1 - p̂ᵢ)^γ (ablation) | AFR · Waterbirds | test worst-group accuracy: 90.4_{\pm1.1} -> 89.3_{\pm1.9} | No clear effect |
Closest papers by meaning
- The Group Robustness is in the Details: Revisiting Finetuning under Spurious Correlations (NeurIPS 2024 · similarity 0.87)
- Group-robust Sample Reweighting for Subpopulation Shifts via Influence Functions (ICLR 2025 · similarity 0.87)
- Label-Efficient Group Robustness via Out-of-Distribution Concept Curation (CVPR 2024 · similarity 0.86)
- Improving Generalization via Meta-Learning on Hard Samples (CVPR 2024 · similarity 0.86)
- Improving Group Robustness on Spurious Correlation Requires Preciser Group Inference (ICML 2024 · similarity 0.86)
- Calibrating Multi-modal Representations: A Pursuit of Group Robustness without Annotations (CVPR 2024 · similarity 0.85)
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.
