AI Paper Atlas

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

Hidden in Plain Sight -- Class Competition Focuses Attribution Maps

Nils Philipp Walter, Jilles Vreeken, Jonas Fischer

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Feature attribution methods

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
add Attribution Lens (class-competition softmax over attribution maps)ResNet-50 attribution methods · Quad-ImageNetRA: avg +53.17%Improved
add Attribution Lens to Guided BackpropGBP on ResNet-50 · Quad-ImageNetRA: +144%Improved
add Attribution Lens to ViT attribution methodsViT-base-16 methods · Quad-ImageNetIoU: avg +103.29%Improved
add Attribution Lens to Input x Gradient and Integrated GradientsIxG, IG on ResNet-50 · Quad-ImageNetIoU: 0.20, +0%No clear effect
add Attribution Lens on real multi-object imagesResNet-50 methods · COCORA, IoU, F1: avg +17.83, +5.83, +4.83Improved
add Attribution Lens on CNN insertion testCNN attribution methods · Insertion AUCAUC: 8-11% on averageImproved
add Attribution Lens on transformer insertion testViT attribution methods · Insertion AUCAUC: ~2%Improved
add Attribution Lens to Grad-CAMGradCam · Insertion AUCAUC: +0% -> -6% (ConvNeXT)Mixed
add Attribution Lens to Grad-CAM on DenseNet121GradCam DenseNet121 · COCOIoU: -25%Hurt
add Attribution Lens on ConvNeXtConvNeXt methods · Quad-ImageNetIoU: avg -1.20%Hurt
add Attribution Lens to perturbation-based methodsShapley, LIME, Occlusion etc. · COCORA vs IoU/F1: RA +13 but IoU -1%Mixed
add Attribution Lens before randomization sanity checkGBP, IxG, ViT methods · Cascading randomizationSpearman similarityImproved

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