AI Paper Atlas

Research map / Learning theory and algorithms / Fairness

KINDER: Kernel-based Independence for Fair Representation Learning via Prototype-space Erasure

Abtin Mogharabin, Jiaee Cheong, Alp Toykan Kaplan, Sinan Kalkan

NeurIPS 2026 · Learning theory and algorithms › Fairness · approach: Fairness-constrained classification · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes KINDER, a concept-erasure method that defines debiasing directly via kernel-based statistical independence in prototype space rather than through a specific adversary or probe.

Problem

Concept erasure for fairness typically defines debiasing indirectly through the failure of a chosen adversary, probe, or fairness penalty, tying the erasure target to a particular decoder family or optimization setup.

Approach

Defines erasure of sensitive attributes directly as statistical independence between representations and sensitive attributes, measured with kernels in a prototype space, instead of depending on an adversary's failure.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
apply KINDER post-hoc to BERT embeddingsBERT · BIOSworst-group macro-F1: 70.7 -> 73.9Improved
apply KINDER post-hoc to BERT embeddingsBERT · MOJIaccuracy: 80.1 -> 79.2Hurt
apply KINDER post-hoc to BERT embeddingsBERT · MOJIworst-class TPR gap: 25.6 -> 16.7Improved
keep native end-to-end task head instead of replacing it with a linear probe (ablation)TabM + KINDER · MIMIC-IV classificationaccuracy: 80.6 -> 92.0Improved
apply KINDER end-to-end with native headTabM · MIMIC-IV classificationF1: 39.9 -> 45.5Improved
apply INLP, RLACE or LEACE post-hoc to TabMTabM · MIMIC-IV classificationaccuracy: 91.9 -> 79.5 (INLP)Hurt
apply KINDER to ResNet regressorResNet · MIMIC-IV regressionintersectional worst-group MAE: 5.55 -> 7.01Hurt
apply KINDER to ResNet regressorResNet · MIMIC-IV regressionMAE: 2.59 -> 2.52Improved
apply KINDER to UNet segmentation featuresUNet · FairSeg intersectionalHD95 gap: 30.7 -> 5.4Improved
apply KINDER to FactorCLFactorCL · FairVLMedworst-group intersectional accuracy: 0.0 -> 41.5Improved
apply KINDER to CLIPCLIP · FairVLMedaccuracy: 73.6 -> 71.6Hurt
apply KINDER to I-JEPAI-JEPA · CelebAaccuracy: 91.1 -> 92.2Improved

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