Research map / Label-efficient and robust learning / Noisy labels
Label-Retrieval-Augmented Diffusion Models for Learning from 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 CARD with LRA-diffusion (retrieval labels plus separate f_p encoder) (ablation) | diffusion classifier with CLIP feature · CIFAR-100 70% PMD | accuracy: 23.45 -> 74.58 | Helped |
| use Gaussian diffusion instead of Dirichlet SS-DDPM (ablation) | LRA-diffusion (CLIP) · CIFAR-100 70% PMD | accuracy: 72.24 -> 74.58 | Helped |
| use diffusion instead of linear probing on sampled neighbor labels (ablation) | CLIP feature, Webvision · WebVision | accuracy: 54.84 -> 84.16 | Helped |
| use sampled neighbor labels instead of mean-label target (ablation) | diffusion with CLIP · WebVision | accuracy: 83.96 -> 84.16 | No clear effect |
| swap CLIP encoder for SimCLR encoder (ablation) | LRA-diffusion · CIFAR-100 35% PMD | accuracy: 81.91 -> 61.39 | Hurt |
| linear probing with sampled labels instead of noisy labels (ablation) | SimCLR feature · CIFAR-10 35% PMD | accuracy: 86.9 -> 63.8 | Hurt |
| LRA-diffusion (SimCLR) versus C2D on CIFAR-100 | SimCLR-based method · CIFAR-100 35% PMD | accuracy: 69.28 -> 61.39 | Hurt |
| use mean-label target instead of sampled labels at noise above 55% (ablation) | LRA-diffusion · CIFAR-10 PMD noise | accuracy | Mixed |
Closest papers by meaning
- Directional Label Diffusion Model for Learning from Noisy Labels (CVPR 2025 · similarity 0.91)
- Label-Noise Robust Diffusion Models (ICLR 2024 · similarity 0.91)
- Adaptive Integration of Partial Label Learning and Negative Learning for Enhanced Noisy Label Learning (AAAI 2024 · similarity 0.89)
- Data Augmentation with Diffusion for Open-Set Semi-Supervised Learning (NeurIPS 2024 · similarity 0.88)
- Learning from Noisy Labels via Conditional Distributionally Robust Optimization (NeurIPS 2024 · similarity 0.88)
- Imprecise Label Learning: A Unified Framework for Learning with Various Imprecise Label Configurations (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.