AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

From Linearity to Non-Linearity: How Masked Autoencoders Capture Spatial Correlations

Anthony Bisulco, Rahul Ramesh, Randall Balestriero, Pratik Chaudhari

ICCV 2025 · Architectures and efficiency › Attention and transformers · approach: Token merging and masked pretraining · 2 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
increase masking ratio from 0.1 to 0.9 (ablation)ViT MAE · CIFAR-10 pretraining timetraining time (hours): 10 hours -> 4 hoursImproved
use a 1-layer decoder instead of the best decoder depth (ablation)ViT MAE · CIFAR-10 fine-tuningaccuracy (%): 95.26% accuracy, only 0.30% worse than best modelNo clear effect
fine-tune only the last Transformer block (freeze the rest) (ablation)ViT MAE · CIFAR-10 fine-tuningaccuracy (%): within 2% of full fine-tuning accuracy, nearly 2x speedupNo clear effect
add MAE pretraining versus supervised-only training (ablation)ViT · CIFAR-10accuracy (%): 83-84% plateau without pretrainingImproved
increase number of encoder layers (ablation)ViT MAE · CIFAR-10 linear probelinear-probe accuracyImproved
increase number of decoder layers beyond 2-4 (ablation)ViT MAE · CIFAR-10 linear probelinear-probe accuracy: optimal decoder contains 2-4 layersNo clear effect
increase masking ratio (ablation)ViT MAE · CIFAR-10 linear probelinear-probe accuracyNo clear effect
increase patch size (ablation)ViT MAE · CIFAR-10 linear probelinear-probe accuracyHurt

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