Research map / Label-efficient and robust learning / Noisy labels
Two Wrongs Don’t Make a Right: Combating Confirmation Bias in Learning with Label Noise
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 |
|---|---|---|---|
| Use Robust LR instead of DivideMix | CIFAR-10 and CIFAR-100 synthetic-noise classification · CIFAR-10, 90% symmetric noise | best test accuracy: 76.0 -> 92.8 | Helped |
| Use Robust LR instead of DivideMix | noisy-label classification · Mini-WebVision | top-1 accuracy: 77.32 -> 81.84 | Helped |
| Use Robust LR instead of NCT | noisy-label classification · ANIMAL-10N | accuracy: 84.1 -> 88.5 | Helped |
| Add label refurbishment by dynamically combining noisy labels with pseudo-labels (ablation) | Robust LR · CIFAR-10, 90% symmetric noise | best test accuracy: 73.8 -> 92.8 | Helped |
| Add strong augmentation (ablation) | Robust LR · CIFAR-10, 90% symmetric noise | best test accuracy: 48.7 -> 92.8 | Helped |
| Use dynamic GMM confidence estimation instead of fixed confidence 0.5 (ablation) | Robust LR · CIFAR-10, 90% symmetric noise | best test accuracy: 87.6 -> 92.8 | Helped |
| Add co-training with two interacting models (ablation) | Robust LR · CIFAR-10, 90% symmetric noise | best test accuracy: 82.8 -> 92.8 | Helped |
| Add label refurbishment by dynamically combining noisy labels with pseudo-labels (ablation) | Robust LR · CIFAR-10, 20% symmetric noise | best test accuracy: 96.3 -> 96.5 | No clear effect |
Closest papers by meaning
- Using Trusted Data to Train Deep Networks on Labels Corrupted by Severe Noise (NeurIPS 2018 · similarity 0.91)
- Mitigating Label Noise through Data Ambiguation (AAAI 2024 · similarity 0.90)
- Handling Label Noise via Instance-Level Difficulty Modeling and Dynamic Optimization (NeurIPS 2025 · similarity 0.89)
- Soften to Defend: Towards Adversarial Robustness via Self-Guided Label Refinement (CVPR 2024 · similarity 0.88)
- Learning from Noisy Labels via Conditional Distributionally Robust Optimization (NeurIPS 2024 · similarity 0.88)
- Combating Semantic Contamination in Learning with Label Noise (AAAI 2025 · 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.