AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

Suppressing Non-Semantic Noise in Masked Image Modeling Representations

Martine Hjelkrem-Tan, Marius Aasan, Rwiddhi Chakraborty, Gabriel Y. Arteaga, Changkyu Choi, Adín Ramírez Rivera

CVPR 2026 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Masked autoencoder pretraining · 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
add SI-score suppression to patch embeddingsCAPI ViT-L14 · ECSSD salient segmentationmax F-beta: 72.456 -> 85.219Improved
add SI-score suppression to patch embeddingsMAE ViT-B16 · ECSSD salient segmentationmax F-beta: 79.952 -> 82.094Improved
add SI-score suppression to patch embeddingsDINOv2 ViT-B16 · ECSSD salient segmentationmax F-beta: 71.319 -> 80.633Improved
add SI-score suppression to patch embeddingsDINOv2 ViT-B16 · DUT-OMRON salient segmentationmax F-beta: 45.923 -> 50.610Improved
add SI-score suppression to patch embeddingsDINOv2 ViT-B16 · DUT-OMRON salient segmentationaccuracy: 75.067 -> 71.762Hurt
add SI-score suppression to patch embeddingsDINOv3 ViT-B16 · DUTS salient segmentationmax F-beta: 39.874 -> 47.624Improved
add SI-score suppression to patch embeddingsMAE ViT-B16 · ImageNet kNN classificationtop-1 accuracy: 59.30 -> 60.62Improved
add SI-score suppression to patch embeddingsDINOv3 ViT-B16 · ImageNet kNN classificationtop-1 accuracy: 81.47 -> 81.48No clear effect
add SI-score suppression to patch embeddingsDINO ViT-B16 (non-MIM) · ADE20k and ImageNet kNNmIoU: 21.208 -> 21.208No clear effect
add SI-score suppression to patch embeddingsMAE ViT-B16 · ADE20k kNN segmentationmIoU: 11.882 -> 12.651Improved
add Fermi-window scaling of SI-scores (ablation)SI suppression on CAPI · ImageNet kNN average-pooledtop-1 accuracy: 55.720 -> 56.444Improved
add Fermi-window scaling of SI-scores (ablation)SI suppression on I-JEPA · ImageNet kNN average-pooledtop-1 accuracy: 71.422 -> 71.390No clear effect

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