Research map / Label-efficient and robust learning / Semi-supervised learning
MarginMatch: Improving Semi-Supervised Learning with Pseudo-Margins
Tiberiu Sosea, Cornelia Caragea
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 training-dynamics pseudo-margin filtering to flexible-threshold pseudo-labeling | FlexMatch-style SSL · CIFAR-100, 4 labels/class | error rate: 39.99 -> 36.97 | Improved |
| add training-dynamics pseudo-margin filtering to flexible-threshold pseudo-labeling | FlexMatch-style SSL · CIFAR-100, 25 labels/class | error rate: 26.96 -> 23.71 | Improved |
| add training-dynamics pseudo-margin filtering to flexible-threshold pseudo-labeling | FlexMatch-style SSL · CIFAR-10, all reported label counts | error rate: 5.04 -> 4.91; 5.04 -> 4.73; 4.19 -> 3.98 | Improved |
| add training-dynamics pseudo-margin filtering to flexible-threshold pseudo-labeling | FlexMatch-style SSL · STL-10, 4 labels/class | error rate: 29.15 -> 25.37 | Improved |
| add training-dynamics pseudo-margin filtering to flexible-threshold pseudo-labeling | FlexMatch-style SSL · SVHN, 25 labels/class | error rate: 3.13 -> 3.14 | No clear effect |
| use EMA smoothing δ=0.95 instead of δ=0.997 (ablation) | MarginMatch · CIFAR-100, 4 labels/class | error rate: 37.91 -> 38.13 | Hurt |
| use EMA smoothing δ=0.99 instead of δ=0.997 (ablation) | MarginMatch · CIFAR-100, 4 labels/class | error rate: 37.91 -> 38.05 | Hurt |
| use EMA smoothing δ=0.995 instead of δ=0.997 (ablation) | MarginMatch · CIFAR-100, 4 labels/class | error rate: 37.91 -> 37.92 | No clear effect |
| use EMA smoothing δ=0.999 instead of δ=0.997 (ablation) | MarginMatch · CIFAR-100, 4 labels/class | error rate: 37.91 -> 39.12 | Hurt |
| replace EMA with simple pseudo-margin averaging (δ=1) (ablation) | MarginMatch · CIFAR-100, 4 labels/class | error rate: 37.91 -> 39.72 | Hurt |
| replace pseudo-margin with average confidence (ablation) | MarginMatch · CIFAR-100, 4 labels/class | error rate: 36.97 -> 41.23 | Hurt |
| replace pseudo-margin with average entropy (ablation) | MarginMatch · CIFAR-100, 4 labels/class | error rate: 36.97 -> 45.10 | Hurt |
Later papers that compared with this method
1 later accepted paper reports MarginMatch as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Bi-Level Optimization for Semi-Supervised Learning with Pseudo-Labeling (AAAI 2025 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- FlexMatch — FlexMatch: Boosting Semi-Supervised Learning with Curriculum Pseudo Labeling (NeurIPS 2021) · better on 15 of 15 shared results
- FixMatch — FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence (NeurIPS 2020) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence (NeurIPS 2020 · similarity 0.91)
- ConMatch: Semi-Supervised Learning with Confidence-Guided Consistency Regularization (ECCV 2022 · similarity 0.91)
- 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)
- Rethinking Confidence Scores and Thresholds in Pseudolabeling-based SSL (ICML 2025 · similarity 0.88)
- Label-Efficient Dataset Pruning via Semi-Supervised Pseudo-Labeling (NeurIPS 2026 · 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.
