Research map / Label-efficient and robust learning / Noisy labels
LNL+K: Enhancing Learning with Noisy Labels Through Noise Source Knowledge Integration
Siqi Wang, Bryan Plummer
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 |
|---|---|---|---|
| add noise source knowledge to CRUST (ablation) | CRUST · CIFAR-10 dominant noise 0.8 | test accuracy: 65.79 -> 80.54 | Improved |
| add noise source knowledge to DISC (ablation) | DISC · BBBC036 cell dataset | test accuracy: 40.55 -> 63.32 | Improved |
| apply standard LNL methods (without knowledge) to BBBC036 | LNL methods · BBBC036 | test accuracy: 63.49 -> 40.55 (DISC) | Hurt |
| add noise source knowledge to CRUST in selecting clean samples (ablation) | CRUST · CIFAR-10 dominant noise 0.8 | clean-sample recall: 36.15 -> 99.04 | Improved |
| add noise source knowledge to FINE on asymmetric noise (ablation) | FINE · CIFAR-100 asymmetric noise 0.4 | test accuracy: 65.11 -> 72.87 | Improved |
| add noise source knowledge to CRUST on asymmetric noise (ablation) | CRUST · CIFAR-10 asymmetric noise 0.4 | test accuracy: 89.40 -> 84.96 | Hurt |
| add noise source knowledge to SFT on CIFAR-100 dominant noise 0.8 (ablation) | SFT · CIFAR-100 dominant noise 0.8 | test accuracy: 41.76 -> 37.96 | Hurt |
| add noise source knowledge to FINE on CHAMMI-CP (ablation) | FINE · CHAMMI-CP | test accuracy: 67.27 -> 67.02 | No clear effect |
| use partial or incomplete noise source knowledge instead of none (ablation) | LNL+K methods · CIFAR-100 dominant noise 0.2 | test accuracy | Improved |
| estimate noise source knowledge with DualT instead of using ground truth sources (ablation) | FINE+k · CIFAR-100 asymmetric noise 0.4 | test accuracy: 72.87 -> 62.80 | Hurt |
| add noise source knowledge to UNICON on Clothing1M (ablation) | UNICON · Clothing1M | test accuracy: 74.56 -> 75.13 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- UNICON — UniCon: Combating Label Noise Through Uniform Selection and Contrastive Learning (CVPR 2022) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning with Noisy Labels Revisited: A Study Using Real-World Human Annotations (ICLR 2022 · similarity 0.89)
- CA2C: A Prior-Knowledge-Free Approach for Robust Label Noise Learning via Asymmetric Co-learning and Co-training (ICCV 2025 · similarity 0.89)
- Noise-Aware Generalization: Robustness to In-Domain Noise and Out-of-Domain Generalization (ICLR 2026 · similarity 0.89)
- Revisiting Learning with Noisy Labels: Active Forgetting and Noise Suppression (CVPR 2026 · similarity 0.89)
- Learning with Open-world Noisy Data via Class-independent Margin in Dual Representation Space (AAAI 2025 · similarity 0.89)
- NoiseGPT: Label Noise Detection and Rectification through Probability Curvature (NeurIPS 2024 · 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.
