Research map / Learning theory and algorithms / Fairness
Latent Space Smoothing for Individually Fair Representations
Momchil Peychev, Anian Ruoss, Mislav Balunović, Maximilian Baader, Martin Vechev
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 adversarial latent-space training (LASSI) over naive representation | Naive · CelebA Smiling (Pale_Skin) | certified fairness: 0.6 -> 98.0 | Improved |
| add adversarial latent-space training (LASSI) over naive representation | Naive · CelebA Earrings (Young) | certified fairness: 59.2 -> 98.0 | Improved |
| add data augmentation with generated similar samples | Naive · CelebA Smiling (Pale_Skin) | certified fairness: 0.6 -> 12.2 | Improved |
| add LASSI over DataAug | DataAug · CelebA Smiling (Pale_Skin) | certified fairness: 12.2 -> 98.0 | Improved |
| add LASSI to imbalanced Earrings task | Naive · CelebA Earrings (Pale_Skin) | accuracy: 81.3 -> 85.0 | Improved |
| add LASSI to Smiling task | Naive · CelebA Smiling (Pale_Skin) | accuracy: 86.3 -> 85.9 | No clear effect |
| combine three sensitive attributes instead of one | LASSI · CelebA Smiling | certified fairness: 98.0 -> 86.5 | Hurt |
| increase adversarial weight lambda2 from 0.025 to 0.05 (ablation) | LASSI · CelebA validation Smiling (Pale_Skin) | certified fairness: 66.2 -> 98.1 | Improved |
| increase lambda2 above 0.05 (0.25) (ablation) | LASSI · CelebA validation Smiling (Pale_Skin) | accuracy: 89.7 -> 79.4 | Hurt |
| increase classification weight lambda1 from 0 to 0.01 in transfer (ablation) | Transfer · FairFace Age-2 | accuracy: 66.4 -> 74.9 | Improved |
| train LASSI directly on the downstream task rather than transfer (ablation) | Transfer lambda1=0.01 · FairFace Age-2 | certified fairness: 91.7 -> 95.0 | Improved |
| use alternative perpendicular attribute vector | LASSI · CelebA Smiling (Pale_Skin) | certified fairness: 34.0 -> 98.8 | Improved |
Closest papers by meaning
- Learning Certified Individually Fair Representations (NeurIPS 2020 · similarity 0.89)
- Fair Representation Learning with Controllable High Confidence Guarantees via Adversarial Inference (NeurIPS 2025 · similarity 0.88)
- Fairness without Demographics through Shared Latent Space-Based Debiasing (AAAI 2024 · similarity 0.87)
- On the Fairness ROAD: Robust Optimization for Adversarial Debiasing (ICLR 2024 · similarity 0.86)
- Causal Adversarial Perturbations for Individual Fairness and Robustness in Heterogeneous Data Spaces (AAAI 2024 · similarity 0.86)
- Kernel-based Equalized Odds: A Quantification of Accuracy-Fairness Trade-off in Fair Representation Learning (NeurIPS 2025 · 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.
