Research map / Image generation / Fast diffusion sampling
Guiding Noisy Label Conditional Diffusion Models with Score-based Discriminator Correction
Dat Nguyen Cong, Hieu Tran Bao, Tung Hoang-Thanh
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 SBDC guidance to EDM | EDM · CIFAR-10 symmetric 50% | CW-FID: 38.6 -> 15.6 | Improved |
| add SBDC guidance to EDM | EDM · CIFAR-10 symmetric 80% | CW-FID: 71.7 -> 48.2 | Improved |
| use SBDC instead of TDSM | EDM · CIFAR-10 instance 40% | CW-FID: 18.1 -> 13.6 | Improved |
| use SBDC instead of TDSM | EDM · CIFAR-100 symmetric 40% | CW-FID: 91.13 -> 77.13 | Improved |
| add SBDC guidance to EDM | EDM · CIFAR-10 symmetric 50% | FID: 2.07 -> 2.24 | Hurt |
| use SBDC instead of TDSM | EDM · Clothing1M | FID: 6.75 -> 4.17 | Improved |
| add SiMix augmentation (ablation) | SBDC · CIFAR-10 asymmetric 20% | CW-Den.: 102.95 -> 104.49 | Improved |
| add Pseudo-clean Shuffle on top of SiMix (ablation) | SBDC · CIFAR-10 asymmetric 20% | CW-Den.: 104.49 -> 114.80 | Improved |
| use CL instead of CORES detector (higher recall) (ablation) | SBDC · CIFAR-10 asymmetric | CW-Den.: 104.5 -> 108.3 | Improved |
| restrict guidance to middle sampling steps only (ablation) | SBDC · CIFAR-10 | inference time: 8m35s -> 8m39s | No clear effect |
| use high guidance strength gamma (ablation) | SBDC · CIFAR-10 | diversity: gamma set to 0.9 | Hurt |
Closest papers by meaning
- Label-Noise Robust Diffusion Models (ICLR 2024 · similarity 0.91)
- Directional Label Diffusion Model for Learning from Noisy Labels (CVPR 2025 · similarity 0.90)
- Structure-Guided Adversarial Training of Diffusion Models (CVPR 2024 · similarity 0.89)
- Learning Diffusion Models with Flexible Representation Guidance (NeurIPS 2025 · similarity 0.89)
- Elucidating the design space of classifier-guided diffusion generation (ICLR 2024 · similarity 0.89)
- Identifying and Solving Conditional Image Leakage in Image-to-Video Diffusion Model (NeurIPS 2024 · 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.
