Research map / Label-efficient and robust learning / Recognition and long-tail
Mitigating Source Bias for Fairer Weak Supervision
Changho Shin, Sonia Cromp, Dyah Adila, Frederic Sala
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 SBM (OT-S) source bias mitigation to labeling functions | weak supervision pipeline · Bank Marketing | accuracy: 0.674 -> 0.847 | Improved |
| Apply SBM (OT-S) source bias mitigation | weak supervision pipeline · Bank Marketing | demographic parity gap: 0.543 -> 0.122 | Improved |
| Apply SBM (OT-S) source bias mitigation | weak supervision pipeline · CivilComments | demographic parity gap: 0.560 -> 0.028 | Improved |
| Apply SBM (OT-S) source bias mitigation | weak supervision pipeline · HateXplain | demographic parity gap: 0.170 -> 0.072 | Improved |
| Apply SBM (OT-S) source bias mitigation | weak supervision pipeline · CelebA | demographic parity gap: 0.308 -> 0.306 | No clear effect |
| Apply SBM to CivilComments with highly unbalanced classes | weak supervision pipeline · CivilComments | F1: 0.223 -> 0.047 | Hurt |
| Use linear OT map (OT-L) for the group mapping on Adult | SBM · Adult | accuracy: 0.717 -> 0.560 | Hurt |
| Embed tabular features as text with LIFT and BERT for nearest neighbor search | SBM (OT-L) · Adult | demographic parity gap: 0.893 -> 0.015 | Improved |
| Embed tabular features as text with LIFT | SBM (OT-S) · Adult | accuracy: 0.723 -> 0.782 | Improved |
| Add LIFT embedding on Bank Marketing | SBM (OT-S) · Bank Marketing | accuracy: 0.847 -> 0.698 | Hurt |
| Add optimal threshold DP post-processing to both methods | SBM (OT-S) · HateXplain | accuracy: 0.539 -> 0.607 | Improved |
| Add optimal threshold DP post-processing to SBM (OT-S) | SBM (OT-S) · HateXplain | accuracy: 0.612 -> 0.607 | Hurt |
Closest papers by meaning
- Unlocking Fairness: a Trade-off Revisited (NeurIPS 2019 · similarity 0.90)
- Mitigating Label Bias in Machine Learning: Fairness through Confident Learning (AAAI 2024 · similarity 0.87)
- Improving Fair Training under Correlation Shifts (ICML 2023 · similarity 0.87)
- Weak Supervision Performance Evaluation via Partial Identification (NeurIPS 2024 · similarity 0.87)
- Causal Fairness under Unobserved Confounding: A Neural Sensitivity Framework (ICLR 2024 · similarity 0.86)
- Navigating Towards Fairness with Data Selection (AAAI 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.
