Research map / Label-efficient and robust learning / Noisy labels
Robustness of Accuracy Metric and its Inspirations in Learning with Noisy Labels
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 |
|---|---|---|---|
| Select the teacher and student using noisy validation accuracy instead of using the last epoch | CE on Clothing1M · Clothing1M, noisy validation | test accuracy: 68.05 -> 70.21 | Helped |
| Select the teacher and student using noisy validation accuracy instead of using the last epoch | Co-T on CIFAR-10 · CIFAR-10, uniform noise 0.4 | test accuracy: 92.18±0.36 -> 94.73±0.20 | Helped |
| Use noisy validation rather than clean validation to select models | NTS on Clothing1M · Clothing1M | test accuracy: 70.02 vs 70.21 | No clear effect |
| Use standard augmentation (ablation) | CE NS on CIFAR-100 · CIFAR-100, asymmetric noise 0.4 | test accuracy: 42.31 -> 59.27 | Helped |
| Replace standard augmentation with strong augmentation (ablation) | CE NS on CIFAR-100 · CIFAR-100, asymmetric noise 0.4 | test accuracy: 59.27 -> 55.61 | Hurt |
| Use strong augmentation (ablation) | CE NS on CIFAR-10 · CIFAR-10, uniform noise 0.4 | test accuracy: 75.04 -> 93.76 | Helped |
| Use strong augmentation (ablation) | CE NS on CIFAR-10 · CIFAR-10, asymmetric noise 0.4 | test accuracy: 76.69 -> 91.49 | Helped |
| Use strong augmentation (ablation) | CE NS on CIFAR-100 · CIFAR-100, uniform noise 0.4 | test accuracy: 41.72 -> 71.71 | Helped |
Closest papers by meaning
- Error-Bounded Correction of Noisy Labels (ICML 2020 · similarity 0.90)
- Learning from Noisy Labels with No Change to the Training Process (ICML 2021 · similarity 0.89)
- Enhanced Sample Selection with Confidence Tracking: Identifying Correctly Labeled Yet Hard-to-Learn Samples in Noisy Data (AAAI 2025 · similarity 0.88)
- Label Noise: Ignorance Is Bliss (NeurIPS 2024 · similarity 0.87)
- Information-theoretic Limits of Online Classification with Noisy Labels (NeurIPS 2024 · similarity 0.87)
- Leveraging Data Symmetries to Select an Optimal Subset of Training Data under Label Noise (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.