Keep It on a Leash: Controllable Pseudo-label Generation Towards Realistic Long-Tailed Semi-Supervised Learning

2025/10/05 by Yaxin Hou, Hou, Yaxin, Bo Han +7
Computer Science · #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning and Data Classification #Text and Document Classification Technologies

paper · pdf · doi:10.48550/arxiv.2510.03993

openalex publication_date 2025/10/05 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Current long-tailed semi-supervised learning methods assume that labeled data exhibit a long-tailed distribution, and unlabeled data adhere to a typical predefined distribution (i.e., long-tailed, uniform, or inverse long-tailed). However, the distribution of the unlabeled data is generally unknown and may follow an arbitrary distribution. To tackle this challenge, we propose a Controllable Pseudo-label Generation (CPG) framework, expanding the labeled dataset with the progressively identified reliable pseudo-labels from the unlabeled dataset and training the model on the updated labeled dataset with a known distribution, making it unaffected by the unlabeled data distribution. Specifically, CPG operates through a controllable self-reinforcing optimization cycle: (i) at each training step, our dynamic controllable filtering mechanism selectively incorporates reliable pseudo-labels from the unlabeled dataset into the labeled dataset, ensuring that the updated labeled dataset follows a known distribution; (ii) we then construct a Bayes-optimal classifier using logit adjustment based on the updated labeled data distribution; (iii) this improved classifier subsequently helps identify more reliable pseudo-labels in the next training step. We further theoretically prove that this optimization cycle can significantly reduce the generalization error under some conditions. Additionally, we propose a class-aware adaptive augmentation module to further improve the representation of minority classes, and an auxiliary branch to maximize data utilization by leveraging all labeled and unlabeled samples. Comprehensive evaluations on various commonly used benchmark datasets show that CPG achieves consistent improvements, surpassing state-of-the-art methods by up to 15.97% in accuracy. The code is available at https://github.com/yaxinhou/CPG.

Citations

• A Square Peg in a Square Hole: Meta-Expert for Long-Tailed Semi-Supervised Learning
• Striving for Simplicity: Simple Yet Effective Prior-Aware Pseudo-Labeling for Semi-Supervised Ultrasound Image Segmentation
• Pursuing Better Decision Boundaries for Long-Tailed Object Detection via Category Information Amount
• Long-Tailed Out-of-Distribution Detection: Prioritizing Attention to Tail
• CDMAD: Class-Distribution-Mismatch-Aware Debiasing for Class-Imbalanced Semi-Supervised Learning
• LTGC: Long-tail Recognition via Leveraging LLMs-driven Generated Content
• BaCon: Boosting Imbalanced Semi-supervised Learning via Balanced Feature-Level Contrastive Learning
• Gradient Reweighting: Towards Imbalanced Class-Incremental Learning
• SimPro: A Simple Probabilistic Framework Towards Realistic Long-Tailed Semi-Supervised Learning
• Three Heads Are Better Than One: Complementary Experts for Long-Tailed Semi-supervised Learning
• How Re-sampling Helps for Long-Tail Learning?
• SemiReward: A General Reward Model for Semi-supervised Learning
• MDCS: More Diverse Experts with Consistency Self-distillation for Long-tailed Recognition
• Flexible Distribution Alignment: Towards Long-tailed Semi-supervised Learning with Proper Calibration
• Balancing Logit Variation for Long-tailed Semantic Segmentation
• SoftMatch: Addressing the Quantity-Quality Trade-off in Semi-supervised Learning
• Balanced Product of Calibrated Experts for Long-Tailed Recognition
• FreeMatch: Self-adaptive Thresholding for Semi-supervised Learning
• CoSSL: Co-Learning of Representation and Classifier for Imbalanced Semi-Supervised Learning
• The Majority Can Help The Minority: Context-rich Minority Oversampling for Long-tailed Classification
• ABC: Auxiliary Balanced Classifier for Class-imbalanced Semi-supervised\n Learning
• FlexMatch: Boosting Semi-Supervised Learning with Curriculum Pseudo Labeling
• HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units
• DASO: Distribution-Aware Semantics-Oriented Pseudo-label for Imbalanced Semi-Supervised Learning
• CReST: A Class-Rebalancing Self-Training Framework for Imbalanced Semi-Supervised Learning
• In Defense of Pseudo-Labeling: An Uncertainty-Aware Pseudo-label Selection Framework for Semi-Supervised Learning
• Long-tailed Recognition by Routing Diverse Distribution-Aware Experts
• Long-tail learning via logit adjustment
• FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence
• Virtual Adversarial Training: A Regularization Method for Supervised and Semi-Supervised Learning
• SGDR: Stochastic Gradient Descent with Warm Restarts
• Wide Residual Networks
• A vector-contraction inequality for Rademacher complexities
• Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning
• Dual Student: Breaking the Limits of the Teacher in Semi-supervised Learning

Related

• Why the pseudo label based semi-supervised learning algorithm is effective?
• Semi-supervised learning by selective training with pseudo labels via confidence estimation
• Smooth Pseudo-Labeling
• In all LikelihoodS: How to Reliably Select Pseudo-Labeled Data for Self-Training in Semi-Supervised Learning
• Three Heads Are Better Than One: Complementary Experts for Long-Tailed Semi-supervised Learning