Research map / Label-efficient and robust learning / Recognition and long-tail
Transferring Fairness under Distribution Shifts via Fair Consistency Regularization
Bang An, Zora Che, Mucong Ding, Furong Huang
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 fair consistency regularization to Laftr self-training | Laftr+FixMatch · FairFace (target) | V_acc: 8.21 -> 2.82 | Improved |
| Add fair consistency regularization to Laftr self-training | Laftr+FixMatch · FairFace (target) | equalized odds: 9.32 -> 5.70 | Improved |
| Add fair consistency regularization to Laftr self-training | Laftr+FixMatch · FairFace (target) | accuracy: 83.87 -> 85.52 | Improved |
| Remove consistency training on target data (ablation) | Ours (w/ Laftr) · FairFace (target) | accuracy: 85.52 -> 82.43 | Hurt |
| Remove consistency training on target data (ablation) | Ours (w/ Laftr) · FairFace (target) | V_acc: 2.82 -> 6.80 | Hurt |
| Remove consistency training on source data (ablation) | Ours (w/ Laftr) · FairFace (target) | equalized odds: 5.70 -> 8.18 | Hurt |
| Use equal group weights instead of dynamic weights (ablation) | Ours (w/ Laftr) · FairFace (target) | V_acc: 2.82 -> 6.86 | Hurt |
| Keep the teacher classifier fixed instead of updating it (ablation) | Ours (w/ Laftr) · FairFace (target) | accuracy: 85.52 -> 79.13 | Hurt |
| Use weaker cropping-and-flipping transformations instead of RandAugment (ablation) | Ours (w/ Laftr) · FairFace (target) | accuracy: 85.52 -> 77.32 | Hurt |
| Use standard rather than group-balanced consistency regularization | Laftr self-training · FairFace (target) | V_acc: 2.82 -> 8.21 | Hurt |
| Apply DANN domain adaptation | Laftr · FairFace (target) | accuracy: 74.35 -> 74.28 | No clear effect |
Closest papers by meaning
- Domain Adaptation meets Individual Fairness. And they get along. (NeurIPS 2022 · similarity 0.93)
- AIM-Fair: Advancing Algorithmic Fairness via Selectively Fine-Tuning Biased Models with Contextual Synthetic Data (CVPR 2025 · similarity 0.88)
- Fairness without Harm: An Influence-Guided Active Sampling Approach (NeurIPS 2024 · similarity 0.88)
- Achievable Fairness on Your Data With Utility Guarantees (NeurIPS 2024 · similarity 0.87)
- Towards Harmless Rawlsian Fairness Regardless of Demographic Prior (NeurIPS 2024 · similarity 0.87)
- FairNet: Dynamic Fairness Correction without Performance Loss via Contrastive Conditional LoRA (NeurIPS 2025 · similarity 0.87)
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.
