Research map / Image generation / Fast diffusion sampling
Refining Generative Process with Discriminator Guidance in Score-based Diffusion Models
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 |
|---|---|---|---|
| Add discriminator guidance to EDM | EDM · CIFAR-10 unconditional | FID: 2.03 -> 1.77 | Helped |
| Add discriminator guidance to EDM | EDM · CIFAR-10 conditional | FID: 1.82 -> 1.64 | Helped |
| Add discriminator guidance to LSGM | LSGM · CIFAR-10 | FID: 2.10 -> 1.94 | Helped |
| Add discriminator guidance to Soft Truncation | Soft Truncation · CelebA 64x64 | FID: 1.90 -> 1.34 | Helped |
| Add discriminator guidance to EDM | EDM · FFHQ 64x64 | FID: 2.39 -> 1.98 | Helped |
| Add discriminator guidance to DiT-XL/2-G | DiT-XL/2 · ImageNet 256x256 | FID: 2.27 -> 1.83 | Helped |
| Add discriminator guidance to DiT-XL/2-G | DiT-XL/2 · ImageNet 256x256 | Recall: 0.57 -> 0.64 | Helped |
| Add discriminator guidance to DiT-XL/2-G | DiT-XL/2 · ImageNet 256x256 | Precision: 0.83 -> 0.78 | Hurt |
| Replace BCE discriminator loss with UKL (ablation) | EDM-G++ · CIFAR-10 | FID: 1.77 -> 1.84 | Hurt |
| Replace BCE discriminator loss with LSIF (ablation) | EDM-G++ · CIFAR-10 | FID: 1.77 -> 1.84 | Hurt |
| Reduce sampling NFE (ablation) | Discriminator Guidance · CIFAR-10 | FID | Hurt |
| Train the discriminator (ablation) | Discriminator Guidance · ImageNet 256x256 | Recall | Helped |
Closest papers by meaning
- Neural Network-Based Score Estimation in Diffusion Models: Optimization and Generalization (ICLR 2024 · similarity 0.89)
- Revisiting Diffusion Models: From Generative Pre-training to One-Step Generation (ICML 2025 · similarity 0.88)
- Training Diffusion-based Generative Models with Limited Data (ICML 2025 · similarity 0.88)
- Unlocking Dataset Distillation with Diffusion Models (NeurIPS 2025 · similarity 0.88)
- Elucidating the design space of classifier-guided diffusion generation (ICLR 2024 · similarity 0.87)
- Learning Diffusion Models with Flexible Representation Guidance (NeurIPS 2025 · 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.