mlsanityData to 14 Oct 2026

Re-Evaluating the Impact of Unseen-Class Unlabeled Data on Semi-Supervised Learning Model

arXiv 2503.00884 ↗First posted 2 Mar 2025cs.LGcited by 3

Not on the map. Why?

Abstract (arXiv)

Semi-supervised learning (SSL) effectively leverages unlabeled data and has been proven successful across various fields. Current safe SSL methods believe that unseen classes in unlabeled data harm the performance of SSL models. However, previous methods for assessing the impact of unseen classes on SSL model performance are flawed. They fix the size of the unlabeled dataset and adjust the proportion of unseen classes within the unlabeled data to assess the impact. This process contravenes the principle of controlling variables. Adjusting the proportion of unseen classes in unlabeled data alters the proportion of seen classes, meaning the decreased classification performance of seen classes may not be due to an increase in unseen class samples in the unlabeled data, but rather a decrease in seen class samples. Thus, the prior flawed assessment standard that ``unseen classes in unlabeled data can damage SSL model performance" may not always hold true. This paper strictly adheres to the principle of controlling variables, maintaining the proportion of seen classes in unlabeled data while only changing the unseen classes across five critical dimensions, to investigate their impact on SSL models from global robustness and local robustness. Experiments demonstrate that unseen classes in unlabeled data do not necessarily impair the performance of SSL models; in fact, under certain conditions, unseen classes may even enhance them.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 26 of its 37 references that are in the map. First posted on arXiv 2 Mar 2025.

Builds on

26 of 37 references.

References · 26

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2017mixup: Beyond Empirical Risk Minimizationcited by 5,183
  3. 2020FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidencecited by 2,085
  4. 2019MixMatch: A Holistic Approach to Semi-Supervised Learningcited by 1,391
  5. 2019Unsupervised Data Augmentation for Consistency Trainingcited by 1,060
  6. 2016Temporal Ensembling for Semi-Supervised Learningcited by 1,057
  7. 2017Virtual Adversarial Training: A Regularization Method for Supervised and Semi-Supervised Learningcited by 920
  8. 2021FlexMatch: Boosting Semi-Supervised Learning with Curriculum Pseudo Labelingcited by 540
  9. 2016Regularization With Stochastic Transformations and Perturbations for Deep Semi-Supervised Learningcited by 441
  10. 2019Interpolation Consistency Training for Semi-Supervised Learningcited by 251
  11. 2021A Survey on Deep Semi-supervised Learningcited by 251
  12. 2022FreeMatch: Self-adaptive Thresholding for Semi-supervised Learningcited by 232
14 more
  1. 2020CoMatch: Semi-supervised Learning with Contrastive Graph Regularizationcited by 167
  2. 2021Open-World Semi-Supervised Learningcited by 165
  3. 2023SoftMatch: Addressing the Quantity-Quality Trade-off in Semi-supervised Learningcited by 133
  4. 2022SimMatch: Semi-supervised Learning with Similarity Matchingcited by 118
  5. 2022USB: A Unified Semi-supervised Learning Benchmark for Classificationcited by 95
  6. 2020Multi-Task Curriculum Framework for Open-Set Semi-Supervised Learningcited by 82
  7. 2023IOMatch: Simplifying Open-Set Semi-Supervised Learning with Joint Inliers and Outliers Utilizationcited by 23
  8. 2023A Graph-Theoretic Framework for Understanding Open-World Semi-Supervised Learningcited by 22
  9. 2023FlatMatch: Bridging Labeled Data and Unlabeled Data with Cross-Sharpness for Semi-Supervised Learningcited by 20
  10. 2022Imbalanced Semi-supervised Learning with Bias Adaptive Classifiercited by 18
  11. 2023Shrinking Class Space for Enhanced Certainty in Semi-Supervised Learningcited by 17
  12. 2022Fix-A-Step: Semi-supervised Learning from Uncurated Unlabeled Datacited by 11
  13. 2024Robust Semi-Supervised Learning in Open Environmentscited by 8
  14. 2023Out-of-Distributed Semantic Pruning for Robust Semi-Supervised Learningcited by 7

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026USE: Uncertainty Structure Estimation for Robust Semi-Supervised Learningcited by 1
  2. 2026Learning to Evaluate: Cost-Effective Model Evaluation on Unlabeled Data with Meta-Learningcited by 1
  3. 2026C-Score: Beyond Accuracy for Robustness Assessment in Semi-Supervised Learning under Open-World Unlabeled Contaminationcited by 0