AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

GRADIEND: Feature Learning within Neural Networks Exemplified through Biases

Jonathan Drechsel, Steffen Herbold

ICLR 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias · 1 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
combine GRADIEND Female/Male with INLPGRADIEND Female/Male · Gender debiasingmean proportional rank: 0.46 -> 0.88Improved
add GRADIEND Female/Male weight modification to INLPINLP · Gender debiasingmean proportional rank: 0.67 -> 0.88Improved
add GRADIEND Female/Male to SentDebiasSentDebias · Gender debiasingmean proportional rank: 0.60 -> 0.64Improved
add INLP to CDACDA · Gender debiasingmean proportional rank: 0.44 -> 0.75Improved
add INLP to DropoutDropout · Gender debiasingmean proportional rank: 0.30 -> 0.71Improved
apply GRADIEND Female/Male to the base modelbase models · Gender debiasingmean proportional rank: 0.17 -> 0.46Improved
apply GRADIEND Asian/White to base modelbase models · Race debiasingmean proportional rank: 0.44 -> 0.58Improved
apply GRADIEND Black/White to base modelbase models · Race debiasingmean proportional rank: 0.44 -> 0.36Hurt
apply GRADIEND Asian/White to base modelbase models · Race debiasingLMS (%): 87.46 -> 82.00Hurt
apply GRADIEND Black/White to base modelbase models · Race debiasingLMS (%): 87.46 -> 87.37No clear effect
apply GRADIEND Christian/Muslim to base modelbase models · Religion debiasingLMS (%): 87.46 -> 84.76Hurt
apply GRADIEND Female/Male to base modelbase models · Gender debiasingLMS (%): 87.46 -> 86.72No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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