Research map / Label-efficient and robust learning / Noisy labels
Learning with Noisy Labels Revisited: A Study Using Real-World Human Annotations
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 |
|---|---|---|---|
| Train with real human annotations instead of synthetic class-dependent noise | most evaluated methods · CIFAR-10N | test accuracy (%): 4.35; 6.01; 4.50; 5.82; 9.00 | Hurt |
| Train with real human annotations instead of synthetic class-dependent noise | ELR · CIFAR-10N | test accuracy (%): -0.78; -0.81; -0.97; -0.51; -1.64 | Helped |
| Train with real human annotations instead of synthetic class-dependent noise | most evaluated methods · CIFAR-100N Noisy | test accuracy (%): 1.20; -0.14; -0.61; -0.85; 1.05; 1.31; 0.65; 1.26; -0.48; -0.72; 0.47 | Mixed |
| Use human annotation aggregation rather than one randomly selected label | CIFAR-10N labels · CIFAR-10N | noise rate: 17.23% -> 9.03% | Helped |
| Use class-dependent synthetic noise instead of feature-dependent human noise | CIFAR-N noise modeling · CIFAR-10N and CIFAR-100N | feature dependence of label noise: p=1.8e^{-36} | Hurt |
| Train on human-noisy rather than synthetic class-dependent labels at the same noise level | CE with ResNet-34 · CIFAR-10N | memorization of wrong-label features | Hurt |
Closest papers by meaning
- Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels (ICML 2020 · similarity 0.91)
- Analysing the Noise Model Error for Realistic Noisy Label Data (AAAI 2021 · similarity 0.91)
- Beyond Class-Conditional Assumption: A Primary Attempt to Combat Instance-Dependent Label Noise (AAAI 2021 · similarity 0.90)
- Unlocking the Power of Open Set: A New Perspective for Open-Set Noisy Label Learning (AAAI 2024 · similarity 0.88)
- Learning with Open-world Noisy Data via Class-independent Margin in Dual Representation Space (AAAI 2025 · similarity 0.88)
- Learning from Noisy Labels via Conditional Distributionally Robust Optimization (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.