XMIX: Combating Extremely Noisy Labels via Local Smoothness in Self-Supervised Feature Space

2026/04/21 by Chengqi Li, Yangdi Lu, Zhihao Shi +3
Computer Science · #Machine Learning and Data Classification #Face and Expression Recognition #Text and Document Classification Technologies

paper · pdf · doi:10.1109/icassp55912.2026.11463541

Abstract

Supervised deep learning models rely on large, accurately labeled datasets, yet noisy annotations are often unavoidable and can severely degrade performance under high noise levels. Recent state-of-the-art methods tackle this by using sample selection strategies that exploit the memorization effect to filter out clean data for semi-supervised learning. However, these methods struggle with extreme noise, class imbalance, and require careful tuning or prior noise knowledge. To overcome these limitations, we propose XMix, which exploits local smoothness in the self-supervised feature space to strengthen sample selection. XMix estimates noise rates via maximum likelihood among feature neighbors, expands and balances the clean set using label-consistent neighbors, and generates more reliable pseudo-labels during semi-supervised learning. Our empirical results show that XMix substantially outperforms existing methods in extremely noisy environments and maintains superior performance in standard LNL benchmarks.

Citations

• PLReMix: Combating Noisy Labels with Pseudo-Label Relaxed Contrastive Representation Learning
• DINOv2: Learning Robust Visual Features without Supervision
• SoftMatch: Addressing the Quantity-Quality Trade-off in Semi-supervised Learning
• PADDLES: Phase-Amplitude Spectrum Disentangled Early Stopping for Learning with Noisy Labels
• CrossSplit: Mitigating Label Noise Memorization through Data Splitting
• ProMix: Combating Label Noise via Maximizing Clean Sample Utility
• SELC: Self-Ensemble Label Correction Improves Learning with Noisy Labels
• UNICON: Combating Label Noise Through Uniform Selection and Contrastive Learning
• L2B: Learning to Bootstrap Robust Models for Combating Label Noise
• Learning with Noisy Labels Revisited: A Study Using Real-World Human Annotations
• Understanding and Improving Early Stopping for Learning with Noisy Labels
• Learning with Feature-Dependent Label Noise: A Progressive Approach
• Understanding deep learning (still) requires rethinking generalization
• Deep Learning with Label Differential Privacy
• Provably End-to-end Label-Noise Learning without Anchor Points
• Multi-Objective Interpolation Training for Robustness to Label Noise
• Learning with Instance-Dependent Label Noise: A Sample Sieve Approach
• Early-Learning Regularization Prevents Memorization of Noisy Labels
• Normalized Loss Functions for Deep Learning with Noisy Labels
• Dual T: Reducing Estimation Error for Transition Matrix in Label-noise Learning
• Combating noisy labels by agreement: A joint training method with co-regularization
• A Simple Framework for Contrastive Learning of Visual Representations
• FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence
• Are Anchor Points Really Indispensable in Label-Noise Learning?
• MixMatch: A Holistic Approach to Semi-Supervised Learning
• Co-teaching: Robust Training of Deep Neural Networks with Extremely Noisy Labels
• Using Trusted Data to Train Deep Networks on Labels Corrupted by Severe Noise
• WebVision Database: Visual Learning and Understanding from Web Data
• Understanding deep learning requires rethinking generalization
• Making Deep Neural Networks Robust to Label Noise: a Loss Correction Approach
• Discrete Distribution Estimation under Local Privacy

Related

• UNICON: Combating Label Noise Through Uniform Selection and Contrastive Learning
• Robust Feature Learning Against Noisy Labels
• Mitigating Memorization in Sample Selection for Learning with Noisy Labels
• Mitigating Instance-Dependent Label Noise: Integrating Self-Supervised Pretraining with Pseudo-Label Refinement
• ProMix: Combating Label Noise via Maximizing Clean Sample Utility