Research map / Label-efficient and robust learning / Semi-supervised learning
Curriculum Labeling: Revisiting Pseudo-Labeling for Semi-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 |
|---|---|---|---|
| Use curriculum self-pacing instead of a fixed threshold (ablation) | Pseudo-labeling with heavy augmentation · CIFAR-10 | test error: 6.21 -> 5.27 | Helped |
| Use curriculum self-pacing instead of fixed thresholding (ablation) | Pseudo-labeling with moderate augmentation · CIFAR-10 | test error: 17.11 -> 8.92 | Helped |
| Reinitialize model parameters after each self-training round rather than fine-tuning (ablation) | Curriculum labeling · CIFAR-10, iteration 5 | test error: 6.42 -> 5.41 | Helped |
| Add Mixup (ablation) | Pseudo-labeling without curriculum or augmentation · CIFAR-10 | top-1 error: 38.47 -> 34.03 | Helped |
| Add SWA without Mixup (ablation) | Pseudo-labeling without curriculum or augmentation · CIFAR-10 | top-1 error: 38.47 -> 37.35 | Helped |
| Add Mixup and SWA (ablation) | Pseudo-labeling without curriculum or augmentation · CIFAR-10 | top-1 error: 38.47 -> 32.85 | Helped |
| Add heavy Random Augmentation, Mixup, and SWA (ablation) | Pseudo-labeling without curriculum · CIFAR-10 | top-1 error: 38.47 -> 8.59 | Helped |
| Use curriculum labeling instead of vanilla pseudo-labeling with no curriculum (ablation) | Pseudo-labeling · CIFAR-10 | test error: 8.59 -> 5.27 | Helped |
| Use a smaller stepping percentile of 5% rather than the standard 20% step (ablation) | Curriculum labeling · CIFAR-10 | test error: 5.27; 5.45 | Mixed |
| Train for more epochs with heavy augmentation and SWA (ablation) | Curriculum labeling · CIFAR-10 | test error: 8.42; 6.46; 7.42 | Mixed |
| Omit SWA while using heavy augmentation (ablation) | Curriculum labeling · CIFAR-10 | test error: 5.41 -> 5.76 | Hurt |
| Add Mixup to moderate-augmentation pseudo-labeling (ablation) | Pseudo-labeling without curriculum · CIFAR-10 | top-1 error: 22.8 -> 16.11 | Helped |
Closest papers by meaning
- All Labels Are Not Created Equal: Enhancing Semi-Supervision via Label Grouping and Co-Training (CVPR 2021 · similarity 0.92)
- Meta Pseudo Labels (CVPR 2021 · similarity 0.91)
- Keep It on a Leash: Controllable Pseudo-label Generation Towards Realistic Long-Tailed Semi-Supervised Learning (NeurIPS 2025 · similarity 0.89)
- Learning with Partial-Label and Unlabeled Data: A Uniform Treatment for Supervision Redundancy and Insufficiency (ICML 2024 · similarity 0.89)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.88)
- Non-Stationary Predictions May Be More Informative: Exploring Pseudo-Labels with a Two-Phase Pattern of Training Dynamics (ICML 2025 · 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.