Research map / Learning theory and algorithms / Fairness
Fair Infinitesimal Jackknife: Mitigating the Influence of Biased Training Data Points Without Refitting
Prasanna Sattigeri, Soumya Ghosh, Inkit Padhi, Pierre Dognin, Kush R. Varshney
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 |
|---|---|---|---|
| apply Fair-IJ post-hoc to ERM model | BERT linear head · CivilComments (Muslim) | Delta EO: 0.314 -> 0.125 | Improved |
| apply Fair-IJ post-hoc to ERM model | BERT linear head · CivilComments (Muslim) | Delta DP: 0.246 -> 0.011 | Improved |
| apply Fair-IJ post-hoc to ERM model | BERT non-linear head · CivilComments (Muslim) | Delta EO: 0.326 -> 0.126 | Improved |
| apply Fair-IJ post-hoc to ERM model | BERT non-linear head · CivilComments (Muslim) | Delta DP: 0.261 -> 0.008 | Improved |
| Fair-IJ instead of GapReg (lambda=1) | BERT linear head · CivilComments (Muslim) | Delta DP: 0.185 -> 0.011 | Improved |
| Fair-IJ instead of GapReg (lambda=1) | BERT non-linear head · CivilComments (Muslim) | Delta EO: 0.144 -> 0.126 | No clear effect |
| Fair-IJ instead of GapReg (lambda=1) | BERT non-linear head · CivilComments (Muslim) | Delta DP: 0.054 -> 0.008 | Improved |
| apply Fair-IJ to ERM model | BERT non-linear head · CivilComments (Muslim) | balanced accuracy (EO): 59.3 -> 59.8 | Improved |
| apply Fair-IJ to ERM model | BERT linear head · CivilComments (Muslim) | balanced accuracy (DP): 58.4 -> 57.1 | Hurt |
| apply Fair-IJ to ERM model | BERT 4 trigger tokens · CivilComments (Muslim) | Delta EO: 0.360 -> 0.102 | Improved |
| apply Fair-IJ to ERM model | BERT 4 trigger tokens · CivilComments (Muslim) | balanced accuracy (EO): 57.5 -> 56.0 | Hurt |
| apply Fair-IJ to ERM model | BERT 10 trigger tokens · CivilComments (Muslim) | Delta DP: 0.234 -> 0.089 | Improved |
Closest papers by meaning
- Fair and Optimal Classification via Post-Processing (ICML 2023 · similarity 0.89)
- Demystifying the Optimal Fair Classifier in Multi-Class Classification (ICML 2026 · similarity 0.89)
- Fairness Constraints: A Flexible Approach for Fair Classification (JMLR 2019 · similarity 0.89)
- The Fragility of Fairness: Causal Sensitivity Analysis for Fair Machine Learning (NeurIPS 2024 · similarity 0.88)
- Enhancing Group Fairness in Online Settings Using Oblique Decision Forests (ICLR 2024 · similarity 0.87)
- Individual Fairness In Strategic Classification (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.
