AI Paper Atlas

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

Human-like Object Grouping in Self-supervised Vision Transformers

Hossein Adeli, Seoyoung Ahn, Andrew Luo, Mengmi Zhang, Nikolaus Kriegeskorte, Gregory Zelinsky

ECCV 2026 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Invariance-based self-supervised learning · 3 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
use DINO-family self-supervised training instead of supervised training (same ViT-B)ViT B · Two-dot groupinggrouping accuracy (%): 70.6 -> 91.9Improved
add large-scale pretraining data (IN21k vs IN1K)supervised ConvNext B · Two-dot groupinggrouping accuracy (%): 60.0 -> 67.4Improved
add large-scale pretraining data (IN21k vs IN1K)supervised ViT B · Human RT predictionnoise-normalized Spearman r: 0.59 -> 0.59No clear effect
distill DINOv3 ViT into a ConvNextConvNext B · Two-dot groupinggrouping accuracy (%): 67.4 -> 86.7Improved
add Gram matrix distillation from DINOv3 ViT B (ablation)IN21k ViT B · Human RT predictionnoise-normalized Spearman r: 0.59 -> 0.72Improved
add Gram matrix distillation from DINOv3 ViT B (ablation)IN1K ViT B · Human RT predictionnoise-normalized Spearman r: 0.59 -> 0.70Improved
add Gram matrix distillation from DINOv3 ViT B (ablation)supervised ViT/ConvNext B (4 models) · Two-dot groupinggrouping accuracy (%): +8 to +18 pointsImproved
add Gram matrix distillation from DINOv3 ViT B (ablation)supervised ViT/ConvNext B (4 models) · Object-centric AUCAUCImproved
replace Gram distillation with CLS token distillation (ablation)IN21k ViT B · Human RT predictionnoise-normalized Spearman r: 0.59 -> 0.59No clear effect
replace Gram distillation with CLS token distillation (ablation)IN1K ViT B · Human RT predictionnoise-normalized Spearman r: 0.59 -> 0.58No clear effect
add Gram distillation (Transformer vs ConvNext) (ablation)supervised models · Behavioral alignmentnoise-normalized Spearman r: Transformers larger gains than ConvNextMixed
use higher object-centric AUC across models9 vision models · Behavioral alignmentSpearman r: r=0.950Improved

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