Research map / Label-efficient and robust learning / Noisy labels
Noise-Aware Generalization: Robustness to In-Domain Noise and Out-of-Domain Generalization
Siqi Wang, Aoming Liu, Bryan A. Plummer
Read the paperCheck your idea against 137,217 papers
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 DL4ND to ERM | ERM · OfficeHome (60% symmetric noise) | AVG accuracy: 43.2 -> 48.9 | Improved |
| add DL4ND to ERM++ | ERM++ · OfficeHome (60% symmetric noise) | AVG accuracy: 52.7 -> 59.8 | Improved |
| add DL4ND to SAGM | SAGM · PACS (real noise) | AVG accuracy: 90.8 -> 93.1 | Improved |
| add DL4ND to ERM++ | ERM++ · PACS (real noise) | AVG accuracy: 92.9 -> 93.3 | Improved |
| add DL4ND to SAGM | SAGM · DomainNet (40% asymmetric noise) | ID accuracy: 43.6 -> 52.8 | Improved |
| add DL4ND to SAGM | SAGM · TerraIncognita (20% asymmetric noise) | ID accuracy: 78.2 -> 77.1 | Hurt |
| add DL4ND to ERM++ | ERM++ · OfficeHome (20% asymmetric noise) | ID accuracy: 78.5 -> 76.6 | Mixed |
| relabel noisy samples instead of removing them (ablation) | SAGM+SWAD w/ DL4ND · VLCS (real noise) | AVG accuracy: 87.8 -> 90.6 | Improved |
| use cross-domain instead of within-domain comparisons for noise detection (ablation) | SAGM+SWAD w/ DL4ND · VLCS (real noise) | AVG accuracy: 88.2 -> 90.6 | Improved |
| build proxies from small-loss samples only (ablation) | SAGM+SWAD w/ DL4ND · VLCS (real noise) | AVG accuracy: 87.6 -> 90.6 | Improved |
| use cross-domain comparisons (ablation) | ERM++ w/ DL4ND · CHAMMI-CP (real noise) | AVG accuracy: 58.5 -> 58.6 | No clear effect |
| build proxies from small-loss samples only (ablation) | ERM++ w/ DL4ND · CHAMMI-CP (real noise) | AVG accuracy: 56.1 -> 58.6 | Improved |
Closest papers by meaning
- Understanding Domain Generalization: A Noise Robustness Perspective (ICLR 2024 · similarity 0.93)
- Rethinking Multi-domain Generalization with A General Learning Objective (CVPR 2024 · similarity 0.89)
- NoiseGPT: Label Noise Detection and Rectification through Probability Curvature (NeurIPS 2024 · similarity 0.89)
- Handling Label Noise via Instance-Level Difficulty Modeling and Dynamic Optimization (NeurIPS 2025 · similarity 0.88)
- Label Noise: Ignorance Is Bliss (NeurIPS 2024 · similarity 0.88)
- Understanding and Mitigating the Label Noise in Pre-training on Downstream Tasks (ICLR 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.
