Research map / Label-efficient and robust learning / Noisy labels
Learning with Instance-Dependent Label Noise: A Sample Sieve Approach
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 CORES² instead of the strongest listed baseline | ResNet34 · Instance-dependent CIFAR10, ε=0.2, 0.4, 0.6 | test accuracy: 89.33 -> 89.50; 82.70 -> 82.84; 73.73 -> 79.66 | Helped |
| Use CORES² instead of the strongest listed baseline | ResNet34 · Instance-dependent CIFAR100, ε=0.2, 0.4, 0.6 | test accuracy: 59.92 -> 61.25; 45.76 -> 47.81; 33.61 -> 37.85 | Helped |
| Add consistency training on sieved corrupted examples | CORES² · Instance-dependent CIFAR10 and CIFAR100 | test accuracy: 89.50 -> 95.42; 82.84 -> 88.45; 79.66 -> 85.53; 61.25 -> 72.91; 47.81 -> 70.66; 37.85 -> 63.08 | Helped |
| Use CORES² instead of the strongest listed baseline | ResNet50 · Clothing1M | clean test accuracy: 72.46 -> 73.24 | Helped |
| Apply CORES²⋆ instead of the strongest listed baseline | ResNet34 · Symmetric/asymmetric CIFAR10 and CIFAR100 | test accuracy: 92.43 -> 93.76; 88.32 -> 89.78; 93.75 -> 95.18; 92.42 -> 94.67; 66.71 -> 72.22; 56.51 -> 59.16; 70.53 -> 75.19; 65.09 -> | Helped |
| Select clean examples with the confidence-regularized sample sieve rather than CE-based selection | CORES² sample sieve · Instance-dependent label noise | F-score | Helped |
Closest papers by meaning
- A Second-Order Approach to Learning With Instance-Dependent Label Noise (CVPR 2021 · similarity 0.94)
- Tackling Instance-Dependent Label Noise via a Universal Probabilistic Model (AAAI 2021 · similarity 0.93)
- Handling Label Noise via Instance-Level Difficulty Modeling and Dynamic Optimization (NeurIPS 2025 · similarity 0.90)
- On the Role of Label Noise in the Feature Learning Process (ICML 2025 · similarity 0.90)
- Self-cognitive Denoising in the Presence of Multiple Noisy Label Sources (ICML 2024 · similarity 0.88)
- Label Noise: Ignorance Is Bliss (NeurIPS 2024 · 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.