Research map / Label-efficient and robust learning / Noisy labels
How does Disagreement Help Generalization against Label Corruption?
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 |
|---|---|---|---|
| Add disagreement-based filtering to small-loss cross-training | Co-teaching · T-ImageNet Symmetry-50% | averaged/maximal test accuracy: 37.09/37.60 -> 41.19/\textbf{41.77} | Helped |
| Add disagreement-based filtering to small-loss cross-training | Co-teaching · T-ImageNet Symmetry-20% | averaged/maximal test accuracy: 45.60/46.36 -> 47.73/\textbf{48.20} | Helped |
| Add disagreement-based filtering to small-loss cross-training | Co-teaching · T-ImageNet Pair-45% | averaged/maximal test accuracy: 27.41/\textbf{27.82} -> 26.54/26.87 | Hurt |
| Add disagreement-based filtering to small-loss cross-training | Co-teaching · CIFAR-10 + CIFAR-100 open-set noise | averaged/maximal test accuracy: 79.43/79.58 -> 79.28/\textbf{79.74} | Mixed |
| Add disagreement-based filtering to small-loss cross-training | Co-teaching · CIFAR-10 + ImageNet-32 open-set noise | averaged/maximal test accuracy: 79.42/79.60 -> 79.89/\textbf{80.52} | Helped |
| Add disagreement-based filtering to small-loss cross-training | Co-teaching · CIFAR-10 + SVHN open-set noise | averaged/maximal test accuracy: 80.12/80.33 -> 80.62/\textbf{80.95} | Helped |
| Use disagreement-only updates without the small-loss and cross-update strategy | Decoupling · Noisy-label benchmarks | test accuracy | Hurt |
| Add disagreement-based filtering to Co-teaching | Co-teaching · MNIST Pair-45% and Symmetry-50% | test accuracy | Helped |
| Add disagreement-based filtering to Co-teaching | Co-teaching · CIFAR-10 Pair-45% and Symmetry-50% | test accuracy | Helped |
| Add disagreement-based filtering to Co-teaching | Co-teaching · CIFAR-100 Pair-45% and Symmetry-50% | test accuracy | Helped |
| Add disagreement-based filtering to Co-teaching | Co-teaching · NEWS Symmetry-20%, Pair-45%, and Symmetry-50% | test accuracy | Helped |
Closest papers by meaning
- Co-teaching: Robust training of deep neural networks with extremely noisy labels (NeurIPS 2018 · similarity 0.92)
- Combating Noisy Labels by Agreement: A Joint Training Method with Co-Regularization (CVPR 2020 · similarity 0.91)
- Addressing Multi-Label Learning with Partial Labels: From Sample Selection to Label Selection (AAAI 2025 · similarity 0.86)
- Enhancing Robustness of Last Layer Two-Stage Fair Model Corrections (NeurIPS 2024 · similarity 0.86)
- Handling Label Noise via Instance-Level Difficulty Modeling and Dynamic Optimization (NeurIPS 2025 · similarity 0.85)
- Mitigating Label Noise through Data Ambiguation (AAAI 2024 · similarity 0.85)
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.