Research map / Label-efficient and robust learning / Recognition and long-tail
Fair Selective Classification Via Sufficiency
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 sufficiency-based regularizer (conditional MI upper bound) to training | selective classifier on Adult · Adult | area between group precision curves: 0.220 -> 0.021 | Helped |
| add sufficiency-based regularizer to training | selective classifier on Adult · Adult | area under accuracy curve: 0.931 -> 0.887 | Hurt |
| add sufficiency-based regularizer to training | selective classifier on CelebA · CelebA | area between group precision curves: 0.094 -> 0.013 | Helped |
| add sufficiency-based regularizer to training | selective classifier on CheXpert-device · CheXpert-device | area between group precision curves: 0.064 -> 0.031 (DRO 0.080) | Helped |
| add sufficiency-based regularizer to training | selective classifier on Civil Comments · Civil Comments | area between group precision curves: 0.026 -> 0.010 (DRO 0.013) | Helped |
| replace cross-entropy with group DRO | selective classifier on CelebA · CelebA | area under accuracy curve: 0.852 -> 0.965 | Helped |
| add sufficiency-based regularizer to training | selective classifier on CelebA · CelebA | area under accuracy curve: 0.965 (DRO) -> 0.975 | Helped |
| add sufficiency-based regularizer to training | selective classifier on CheXpert-device · CheXpert-device | area under accuracy curve: 0.929 -> 0.934 (DRO 0.933) | No clear effect |
Closest papers by meaning
- On Group Sufficiency Under Label Bias (NeurIPS 2025 · similarity 0.91)
- Selective Classification Can Magnify Disparities Across Groups (ICLR 2021 · similarity 0.91)
- Selective Omniprediction and Fair Abstention (NeurIPS 2025 · similarity 0.89)
- Navigating Towards Fairness with Data Selection (AAAI 2025 · similarity 0.86)
- Accuracy vs. Accuracy: Computational Tradeoffs Between Classification Rates and Utility (NeurIPS 2026 · similarity 0.85)
- Subgroups Matter for Robust Bias Mitigation (ICML 2025 · 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.