Research map / Label-efficient and robust learning / Noisy labels
An accurate detection is not all you need to combat label noise in web-noisy datasets
Paul Albert, Kevin McGuinness, Eric Arazo, Tarun Krishna, Noel O Connor, Jack Valmadre
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| alternate noise detection between PLS and linear-separation metric every epoch (LSA) (ablation) | PLS · CNWL 20% noise | accuracy: 62.30 -> 64.20 | Improved |
| use OR combination of PLS and linear-separation detections (ablation) | PLS-LSA · CNWL 20% noise | accuracy: 62.30 -> 58.79 | Hurt |
| use accurate linear-separation noise detector alone instead of PLS (ablation) | noise-robust training · CNWL 80% noise | accuracy: 48.71 -> 47.11 | Hurt |
| use accurate linear-separation noise detector alone instead of PLS (ablation) | noise-robust training · CNWL 20% noise | accuracy: 62.30 -> 62.97 | Mixed |
| add strong data augmentation to PLS-LSA (ablation) | PLS-LSA · CNWL 80% noise | top-1 accuracy: 43.55 -> 48.20 | Improved |
| add SimCLR initialization to PLS-LSA (ablation) | PLS-LSA · mini-WebVision | top-1 accuracy: 78.68 -> 81.36 | Improved |
| add SimCLR initialization to PLS-LSA (ablation) | PLS-LSA · CNWL 20% noise | top-1 accuracy: 64.04 -> 64.61 | No clear effect |
| add co-training with two networks (PLS-LSA+) | PLS-LSA · CNWL 80% noise | top-1 accuracy: 49.53 -> 52.03 | Improved |
| apply PLS-LSA on fine-grained web datasets with ImageNet init | PLS · Web-Aircraft | top-1 accuracy: 87.58 -> 87.70 | No clear effect |
Closest papers by meaning
- A Noisy Elephant in the Room: Is Your Out-of-Distribution Detector Robust to Label Noise? (CVPR 2024 · similarity 0.87)
- Enhanced Sample Selection with Confidence Tracking: Identifying Correctly Labeled Yet Hard-to-Learn Samples in Noisy Data (AAAI 2025 · similarity 0.85)
- Information-theoretic Limits of Online Classification with Noisy Labels (NeurIPS 2024 · similarity 0.85)
- EAT: Towards Long-Tailed Out-of-Distribution Detection (AAAI 2024 · similarity 0.85)
- How Does Unlabeled Data Provably Help Out-of-Distribution Detection? (ICLR 2024 · similarity 0.85)
- Provable Discriminative Hyperspherical Embedding for Out-of-Distribution Detection (AAAI 2025 · similarity 0.84)
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.
