Research map / Label-efficient and robust learning / Semi-supervised learning
Towards Realistic Long-Tailed Semi-Supervised Learning: Consistency Is All You Need
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 |
|---|---|---|---|
| Use ACR with adaptive pseudo-label refinement and a dual-branch network | FixMatch · CIFAR10-LT and CIFAR100-LT, consistent, uniform, and reversed distributions | top-1 accuracy | Helped |
| Use ACR instead of TRAS | ImageNet-127 · 32×32 and 64×64 | top-1 accuracy: 46.2 -> 57.2; 54.1 -> 63.6 | Helped |
| Add the proposed sample-mask principle rather than standard confidence masking (ablation) | ACR · CIFAR100-LT, consistent, uniform, and reversed | top-1 accuracy: 55.0 -> 55.7; 63.7 -> 66.0; 55.0 -> 57.0 | Helped |
| Add adaptive rather than fixed logit adjustment (ablation) | ACR · CIFAR10-LT and CIFAR100-LT, across consistent, uniform, and reversed distributions | top-1 accuracy: 76.8 vs 81.6; 92.4 vs 92.1; 85.1 vs 85.0; 53.5 vs 55.7; 62.8 vs 66.0; 56.1 vs 57.0 | Mixed |
| Add logit adjustment for the balanced branch (ablation) | ACR · CIFAR10-LT and CIFAR100-LT, across consistent, uniform, and reversed distributions | top-1 accuracy: 74.3 vs 81.6; 90.6 vs 92.1; 83.5 vs 85.0; 54.5 vs 55.7; 66.2 vs 66.0; 56.7 vs 57.0 | Mixed |
| Use balanced softmax rather than standard cross-entropy (ablation) | ACR · CIFAR10-LT, consistent | top-1 accuracy: 76.7 -> 81.6 | Helped |
| Use balanced softmax rather than standard cross-entropy (ablation) | ACR · CIFAR100-LT, consistent | top-1 accuracy: 55.3 vs 55.7 | No clear effect |
| Add gradients from the balanced branch to shared feature learning (ablation) | ACR · CIFAR10-LT and CIFAR100-LT, consistent | top-1 accuracy: 73.7 -> 81.6; 54.3 -> 55.7 | Helped |
| Include labeled examples in the unlabeled training set (ablation) | ACR · CIFAR10-LT, reversed | top-1 accuracy: 79.9 -> 85.0 | Helped |
Closest papers by meaning
- Keep It on a Leash: Controllable Pseudo-label Generation Towards Realistic Long-Tailed Semi-Supervised Learning (NeurIPS 2025 · similarity 0.91)
- Continuous Contrastive Learning for Long-Tailed Semi-Supervised Recognition (NeurIPS 2024 · similarity 0.90)
- A Square Peg in a Square Hole: Meta-Expert for Long-Tailed Semi-Supervised Learning (ICML 2025 · similarity 0.90)
- Three Heads Are Better than One: Complementary Experts for Long-Tailed Semi-supervised Learning (AAAI 2024 · similarity 0.89)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.88)
- SimPro: A Simple Probabilistic Framework Towards Realistic Long-Tailed Semi-Supervised Learning (ICML 2024 · similarity 0.88)
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.