Research map / Label-efficient and robust learning / Noisy labels
FINE Samples for Learning with Noisy Labels
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 |
|---|---|---|---|
| replace loss-based selection with FINE eigenvector alignment selection (ablation) | Co-teaching · CIFAR-10 Sym 80% | test accuracy: 66.3 -> 74.2 | Helped |
| replace Co-teaching selection with FINE (ablation) | Co-teaching · CIFAR-100 Asym 40% | test accuracy: 47.7 -> 64.8 | Helped |
| replace loss-based filtering with FINE (ablation) | DivideMix · CIFAR-10 Sym 90% | test accuracy (last): 75.4 -> 89.6 | Helped |
| replace loss-based filtering with FINE (ablation) | DivideMix · CIFAR-100 Sym 90% | test accuracy (last): 31.0 -> 31.2 | No clear effect |
| replace loss-based filtering with FINE (ablation) | DivideMix · CIFAR-100 Sym 20% | test accuracy (last): 76.9 -> 78.8 | Helped |
| use FINE instead of TopoFilter detector | TopoFilter · CIFAR-10 Sym 80% | test accuracy: 46.8 -> 69.4 | Helped |
| use FINE detector instead of CRUST | CRUST · CIFAR-100 Sym 50% | test accuracy: 62.3 -> 64.2 | Helped |
| use FINE instead of LongReMix on high noise | LongReMix · CIFAR-10 Sym 80% best | test accuracy: 93.9 -> 93.5 | Hurt |
| add FINE to noise-robust losses (GCE/SCE/ELR) (ablation) | noise-robust losses · CIFAR-10/100 severe noise | test accuracy | Helped |
| use FINE in Clothing1M training vs standard | FINE · Clothing1M | test accuracy: 68.94 -> 72.91 | Helped |
| approximate eigenvector with 1% of data (ablation) | FINE · CIFAR-10 Sym 80% | cosine similarity: 0.99 | No clear effect |
| use FINE instead of MCD Mahalanobis estimator | detector · CIFAR | F-score | Helped |
Closest papers by meaning
- Toward Robustness against Label Noise in Training Deep Discriminative Neural Networks (NeurIPS 2017 · similarity 0.90)
- Delving into Noisy Label Detection with Clean Data (ICML 2023 · similarity 0.90)
- Coresets for Robust Training of Deep Neural Networks against Noisy Labels (NeurIPS 2020 · similarity 0.90)
- On the Role of Label Noise in the Feature Learning Process (ICML 2025 · similarity 0.90)
- Distilling Effective Supervision From Severe Label Noise (CVPR 2020 · similarity 0.90)
- Handling Label Noise via Instance-Level Difficulty Modeling and Dynamic Optimization (NeurIPS 2025 · similarity 0.89)
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.