Research map / Label-efficient and robust learning / Semi-supervised learning
Realistic Evaluation of Deep Semi-Supervised Learning Algorithms
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 stochastic consistency regularization | WRN-28-2 · CIFAR-10, 4,000 labels | test error: 20.26 $\pm$ .38\% -> 16.37 $\pm$ .63\% | Helped |
| Add an exponential moving average teacher | WRN-28-2 · CIFAR-10, 4,000 labels | test error: 20.26 $\pm$ .38\% -> 15.87 $\pm$ .28\% | Helped |
| Add virtual adversarial consistency regularization | WRN-28-2 · CIFAR-10, 4,000 labels | test error: 20.26 $\pm$ .38\% -> 13.86 $\pm$ .27\% | Helped |
| Add entropy minimization to VAT | WRN-28-2 with VAT · CIFAR-10, 4,000 labels | test error: 13.86 $\pm$ .27\% -> 13.13 $\pm$ .39\% | Helped |
| Use confident pseudo-labels on unlabeled data | WRN-28-2 · CIFAR-10, 4,000 labels | test error: 20.26 $\pm$ .38\% -> 17.78 $\pm$ .57\% | Helped |
| Add stochastic consistency regularization | WRN-28-2 · SVHN, 1,000 labels | test error: 12.83 $\pm$ .47\% -> 7.19 $\pm$ .27\% | Helped |
| Add an exponential moving average teacher | WRN-28-2 · SVHN, 1,000 labels | test error: 12.83 $\pm$ .47\% -> 5.65 $\pm$ .47\% | Helped |
| Add virtual adversarial consistency regularization | WRN-28-2 · SVHN, 1,000 labels | test error: 12.83 $\pm$ .47\% -> 5.63 $\pm$ .20\% | Helped |
| Add entropy minimization to VAT | WRN-28-2 with VAT · SVHN, 1,000 labels | test error: 5.63 $\pm$ .20\% -> 5.35 $\pm$ .19\% | Helped |
| Use confident pseudo-labels on unlabeled data | WRN-28-2 · SVHN, 1,000 labels | test error: 12.83 $\pm$ .47\% -> 7.62 $\pm$ .29\% | Helped |
| Train a larger supervised model with strong regularization instead of using unlabeled data | CIFAR-10 classifier · CIFAR-10, 4,000 labels | test error: 13.13\% -> 13.4\% | Helped |
| Pretrain on ImageNet and fine-tune on CIFAR-10 | WRN-28-2 · CIFAR-10, 4,000 labels | test error: 13.13\% -> 12.09\% | Helped |
Closest papers by meaning
- Don’t fear the unlabelled: safe semi-supervised learning via debiasing (ICLR 2023 · similarity 0.93)
- Erasing the Bias: Fine-Tuning Foundation Models for Semi-Supervised Learning (ICML 2024 · similarity 0.90)
- InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning (ICML 2024 · similarity 0.89)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.89)
- Revisiting Semi-Supervised Learning in the Era of Foundation Models (NeurIPS 2025 · similarity 0.89)
- Contrastive Credibility Propagation for Reliable Semi-supervised Learning (AAAI 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.