Research map / Learning theory and algorithms / Fairness
Benchmarking Bias Mitigation Toward Fairness Without Harm from Vision to LVLMs
Xuwei Tan, Ziyu Hu, Xueru Zhang
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 RandAugment | ERM · Fitz17k | AUC: 89.74 -> 91.29 | Improved |
| add RandAugment | ERM · FairFace | accuracy: 66.76 -> 68.37 | Improved |
| add RandAugment | ERM · CelebA | accuracy: 86.57 -> 86.72 | No clear effect |
| add RandAugment | ERM · FairFace | accuracy gap: 0.87 -> 1.44 | Hurt |
| add DFR | ERM · Waterbirds | accuracy: 85.63 -> 89.83 | Improved |
| add OxonFair | ERM · Waterbirds | accuracy: 85.63 -> 90.27 | Improved |
| add FSCL | ERM · Waterbirds | accuracy gap: 2.87 -> 1.10 | Improved |
| add MCDP regularization | ERM · CelebA | accuracy: 86.57 -> 80.26 | Hurt |
| add MCDP regularization | ERM · CelebA | EqOdd: 81.91 -> 89.63 | Improved |
| add Decoupled classifier | ERM · Waterbirds | accuracy: 85.63 -> 74.64 | Hurt |
| use zero-shot CLIP instead of tuned ERM | ERM · CelebA | accuracy: 86.57 -> 74.07 | Hurt |
| use zero-shot Qwen2.5-VL 72B instead of tuned ERM | ERM · Waterbirds | accuracy: 85.63 -> 92.41 | Improved |
Closest papers by meaning
- FairCLIP: Harnessing Fairness in Vision-Language Learning (CVPR 2024 · similarity 0.90)
- Subgroups Matter for Robust Bias Mitigation (ICML 2025 · similarity 0.89)
- AIM-Fair: Advancing Algorithmic Fairness via Selectively Fine-Tuning Biased Models with Contextual Synthetic Data (CVPR 2025 · similarity 0.89)
- FairNet: Dynamic Fairness Correction without Performance Loss via Contrastive Conditional LoRA (NeurIPS 2025 · similarity 0.89)
- Interpretable Debiasing of Vision-Language Models for Social Fairness (CVPR 2026 · similarity 0.89)
- Stay Fair! Ensuring Group Fairness in Diffusion Models Across Guidance Scales (NeurIPS 2026 · similarity 0.89)
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.
