Research map / Label-efficient and robust learning / Recognition and long-tail
MetAug: Contrastive Learning via Meta Feature Augmentation
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 margin-injected meta feature augmentation (ablation) | MetAug with OUCL · CIFAR10 | accuracy: 88.79 -> 91.09 | Helped |
| add margin-injected meta feature augmentation (ablation) | MetAug with OUCL · STL-10 | accuracy: 88.31 -> 90.26 | Helped |
| use OUCL alongside meta feature augmentation (ablation) | MetAug · ImageNet (L, ab views) | top-1 accuracy: 64.4 -> 65.1 | Helped |
| add meta feature augmentation to OUCL (ablation) | MetAug · ImageNet (L, ab views) | top-1 accuracy: 63.9 -> 65.1 | Helped |
| add RandAugment (ablation) | MetAug · ImageNet (Y, Db, Dr views) | top-1 accuracy: 66.4 -> 66.7 | Helped |
| use the unattenuated Γ weighting factor instead of its linear attenuation variant (ablation) | MetAug (only OUCL) · Tiny ImageNet | accuracy | Hurt |
| use margin-injected meta feature augmentation (ablation) | MetAug · CIFAR100 | accuracy | Helped |
| evaluate with augmented features instead of the standard encoder representation (ablation) | MetAug · CIFAR10 | accuracy: 85.85 vs 85.48; 85.91 vs 85.99; 86.57 vs 86.65; 87.42 vs 87.41; 87.72 vs 87.87; 87.26 vs 87.47; 86.90 vs 87.19; 86.12 vs | Mixed |
Closest papers by meaning
- The Close Relationship Between Contrastive Learning and Meta-Learning (ICLR 2022 · similarity 0.91)
- A Statistical Theory of Contrastive Learning via Approximate Sufficient Statistics (NeurIPS 2025 · similarity 0.88)
- What Should Not Be Contrastive in Contrastive Learning (ICLR 2021 · similarity 0.88)
- Chaos is a Ladder: A New Theoretical Understanding of Contrastive Learning via Augmentation Overlap (ICLR 2022 · similarity 0.88)
- On the Similarities of Embeddings in Contrastive Learning (ICML 2025 · similarity 0.87)
- DiffAug: Enhance Unsupervised Contrastive Learning with Domain-Knowledge-Free Diffusion-based Data Augmentation (ICML 2024 · 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.