Research map / Probabilistic and scientific modeling / Variational and generative sampling
Understanding Masked Autoencoders via Hierarchical Latent Variable Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| raise masking ratio from 0.1 to 0.75 (ablation) | MAE ViT-Base pretrained on IN1K · ImageNet-1K linear probing | top-1 accuracy: 47.45 to 67.41 | Helped |
| raise masking ratio from 0.75 to 0.9 (ablation) | MAE ViT-Base · ImageNet-1K linear probing | top-1 accuracy: 67.41 to 62.97 | Hurt |
| increase patch size from 8 to 32 at masking ratio 0.75 (ablation) | MAE ViT-Base · ImageNet-1K linear probing | top-1 accuracy: 62.57 to 73.31 | Helped |
| raise masking ratio from 0.1 to 0.75 (ablation) | MAE ViT-Base · COCO object detection (Mask R-CNN) | box AP: 30.47 to 39.72 | Helped |
| raise masking ratio from 0.75 to 0.9 (ablation) | MAE ViT-Base · COCO object detection (Mask R-CNN) | box AP: 39.72 to 37.17 | Hurt |
| raise masking ratio from 0.75 to 0.9 (ablation) | MAE ViT-Base · ImageNet-A robustness | top-1 accuracy: 3.76 to 2.76 | Hurt |
| masking ratio extremes (0.1 and 0.9) versus intermediate ratios (ablation) | MAE ViT-Base · shape bias | shape bias: 0.2563 at 0.75 vs 0.2014 at 0.9 and 0.1352 at 0.1 | Hurt |
| increase patch size from 16 to 32 at masking ratio 0.75 (ablation) | MAE ViT-Base · ImageNet-1K linear probing | top-1 accuracy: 68.96 to 73.31 | Helped |
Closest papers by meaning
- Learning Mask Invariant Mutual Information for Masked Image Modeling (ICLR 2025 · similarity 0.93)
- Self-Guided Masked Autoencoder (NeurIPS 2024 · similarity 0.92)
- Extending Video Masked Autoencoders to 128 frames (NeurIPS 2024 · similarity 0.91)
- Contrastive Tuning: A Little Help to Make Masked Autoencoders Forget (AAAI 2024 · similarity 0.91)
- The Dynamic Duo of Collaborative Masking and Target for Advanced Masked Autoencoder Learning (AAAI 2025 · similarity 0.90)
- R-MAE: Regions Meet Masked Autoencoders (ICLR 2024 · similarity 0.90)
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.