AI Paper Atlas

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

NeurIPS 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Group-robust classification · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
Apply SBM (OT-S) source bias mitigation to labeling functionsweak supervision pipeline · Bank Marketingaccuracy: 0.674 -> 0.847Improved
Apply SBM (OT-S) source bias mitigationweak supervision pipeline · Bank Marketingdemographic parity gap: 0.543 -> 0.122Improved
Apply SBM (OT-S) source bias mitigationweak supervision pipeline · CivilCommentsdemographic parity gap: 0.560 -> 0.028Improved
Apply SBM (OT-S) source bias mitigationweak supervision pipeline · HateXplaindemographic parity gap: 0.170 -> 0.072Improved
Apply SBM (OT-S) source bias mitigationweak supervision pipeline · CelebAdemographic parity gap: 0.308 -> 0.306No clear effect
Apply SBM to CivilComments with highly unbalanced classesweak supervision pipeline · CivilCommentsF1: 0.223 -> 0.047Hurt
Use linear OT map (OT-L) for the group mapping on AdultSBM · Adultaccuracy: 0.717 -> 0.560Hurt
Embed tabular features as text with LIFT and BERT for nearest neighbor searchSBM (OT-L) · Adultdemographic parity gap: 0.893 -> 0.015Improved
Embed tabular features as text with LIFTSBM (OT-S) · Adultaccuracy: 0.723 -> 0.782Improved
Add LIFT embedding on Bank MarketingSBM (OT-S) · Bank Marketingaccuracy: 0.847 -> 0.698Hurt
Add optimal threshold DP post-processing to both methodsSBM (OT-S) · HateXplainaccuracy: 0.539 -> 0.607Improved
Add optimal threshold DP post-processing to SBM (OT-S)SBM (OT-S) · HateXplainaccuracy: 0.612 -> 0.607Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy