Research map / Label-efficient and robust learning / Self-supervised and contrastive
Adversarial Masking for Self-Supervised Learning
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 |
|---|---|---|---|
| Add adversarially learned masks to BYOL | BYOL with ResNet-18 · ImageNet100-S | linear accuracy: 56.3 -> 61.4 | Helped |
| Add adversarially learned masks to SimSiam | SimSiam with ResNet-18 · STL10 | linear accuracy: 84.8 -> 86.4 | Helped |
| Use single-forward-pass ADIOS with BYOL | BYOL with ResNet-18 · ImageNet100 | linear accuracy: 74.3 -> 80.8 | Helped |
| Add ADIOS pretraining | BYOL with ResNet-18 · CIFAR10 transfer, linear evaluation | accuracy: 29.9 -> 39.2 | Helped |
| Add ADIOS pretraining | SimSiam with ResNet-18 · CIFAR100 transfer, fine-tuning | accuracy: 65.0 -> 71.0 | Helped |
| Add ADIOS pretraining | SimCLR with ResNet-18 · Flowers102 transfer, linear evaluation | accuracy: 42.5 -> 50.2 | Helped |
| Add ADIOS to BYOL | BYOL with ResNet-18 · Backgrounds challenge M.R. | accuracy: 45.6 -> 47.8 | Helped |
| Use learned ADIOS masks instead of MAE random masks (ablation) | ADIOS-like masking framework · ImageNet100-S | linear accuracy: 43.7 -> 59.2 | Helped |
| Use learned ADIOS masks instead of BEiT random masks (ablation) | ADIOS-like masking framework · STL10 | linear accuracy: 80.7 -> 86.0 | Helped |
| Use MAE random masks instead of no mask (ablation) | ADIOS-like masking framework · ImageNet100-S | linear accuracy: 57.0 -> 43.7 | Hurt |
| Use BEiT random masks instead of no mask (ablation) | ADIOS-like masking framework · STL10 | linear accuracy: 84.7 -> 80.7 | Hurt |
| Use ground-truth object masks rather than shuffled ground-truth masks (ablation) | ADIOS-like masking framework · CLEVR | F1-macro: 0.332 \std{6e-3} -> 0.373 \std{7e-3} | Helped |
Closest papers by meaning
- Good Helper Is around You: Attention-Driven Masked Image Modeling (AAAI 2023 · similarity 0.89)
- Architecture-Agnostic Masked Image Modeling -- From ViT back to CNN (ICML 2023 · similarity 0.89)
- From Prototypes to General Distributions: An Efficient Curriculum for Masked Image Modeling (CVPR 2025 · similarity 0.89)
- On Data Scaling in Masked Image Modeling (CVPR 2023 · similarity 0.88)
- FSFM: A Generalizable Face Security Foundation Model via Self-Supervised Facial Representation Learning (CVPR 2025 · similarity 0.87)
- Learning Mask Invariant Mutual Information for Masked Image Modeling (ICLR 2025 · similarity 0.87)
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.