Research map / Learning theory and algorithms / Fairness
Chasing Fairness Under Distribution Shift: A Model Weight Perturbation Approach
Zhimeng Jiang, Xiaotian Han, Hongye Jin, Guanchu Wang, Rui Chen, Na Zou, Xia Hu
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 RFR instead of the second-best baseline under mild synthetic shift | RFR · Adult (1.0, 2.0) synthetic shift | demographic parity gap: 47.0% | Improved |
| Use RFR instead of the second-best baseline under mild synthetic shift | RFR · Adult (1.0, 2.0) synthetic shift | equal opportunity gap: 84.8% | Improved |
| Increase synthetic shift intensity from (1.0, 2.0) to (3.0, 6.0) | RFR · Adult synthetic shift | demographic parity gap: 0.44 -> 1.75 | Hurt |
| Use RFR instead of the second-best baseline under temporal shift | RFR · ACS-I 2016->2018 | demographic parity gap: 34.9% | Improved |
| Use RFR instead of the second-best baseline under temporal shift | RFR · ACS-I 2016->2018 | equal opportunity gap: 34.4% | Improved |
| Use RFR instead of ADV under spatial shift MI to CA | RFR · ACS-I | demographic parity gap: 1.04 -> 1.35 | Hurt |
| Increase lambda from 0.0 toward 10.0 (ablation) | RFR · ACS spatial and temporal shift | accuracy and demographic parity | Mixed |
| Vary perturbation radius rho (ablation) | RFR · ACS-I and ACS-E | accuracy and demographic parity | Mixed |
| Remove L_DP term from the RFR objective (ablation) | RFR · ACS-I and ACS-E | demographic parity gap | Improved |
| Remove distribution shift (random split) vs RFR under 2016->2018 shift | RFR · ACS-I | demographic parity gap: 1.36 -> 1.88 | Hurt |
| Use REG instead of RFR without distribution shift | RFR · ACS-E | demographic parity gap: 1.09 -> 1.05 | Improved |
Closest papers by meaning
- fairret: a Framework for Differentiable Fairness Regularization Terms (ICLR 2024 · similarity 0.88)
- Fairness without Harm: An Influence-Guided Active Sampling Approach (NeurIPS 2024 · similarity 0.87)
- Enhancing Robust Fairness via Confusional Spectral Regularization (ICLR 2025 · similarity 0.87)
- Optimal Fair Learning Robust to Adversarial Distribution Shift (ICML 2025 · similarity 0.87)
- A Comprehensive View of Fairness through Distributional Stability (NeurIPS 2026 · similarity 0.86)
- On the Fairness ROAD: Robust Optimization for Adversarial Debiasing (ICLR 2024 · 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.
