Research map / Image generation / Fast diffusion sampling
Guiding a Diffusion Model by Swapping Its Tokens
Weijia Zhang, Yuehao Liu, Shanyan Guan, Wu Ran, Yanhao Ge, Wei Li, Chao Ma
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 SSG guidance (unconditional) | SDXL · MS-COCO 2014 unconditional (30k) | FID: 119.04 -> 70.91 | Improved |
| add SSG guidance (unconditional) | SD1.5 · ImageNet val unconditional (50k) | FID: 74.11 -> 63.05 | Improved |
| add SSG guidance (conditional) | SDXL · MS-COCO 2014 conditional (30k captions) | FID: 45.09 -> 21.73 | Improved |
| add SSG guidance (conditional) | SDXL · MS-COCO 2017 conditional (5k captions) | FID: 54.93 -> 31.92 | Improved |
| swap dissimilar tokens instead of random tokens (ablation) | SSG · MS-COCO 2014 conditional (3k images, ablation) | ImageReward: 0.283 -> 0.297 | Improved |
| swap dissimilar tokens instead of random tokens (ablation) | SSG · MS-COCO 2014 conditional (3k images, ablation) | FID: 32.28 -> 31.41 | Mixed |
| swap dissimilar tokens instead of similar tokens (ablation) | SSG · MS-COCO 2014 conditional (3k images, ablation) | ImageReward: 0.110 -> 0.297 | Improved |
| swap dissimilar tokens instead of similar tokens (ablation) | SSG · MS-COCO 2014 conditional (3k images, ablation) | FID: 28.74 -> 31.41 | Hurt |
| add channel swap to spatial swap (ablation) | SSG · MS-COCO 2014 conditional (3k images, ablation) | ImageReward: 0.272 -> 0.297 | Improved |
| add spatial swap to channel swap (ablation) | SSG · MS-COCO 2014 conditional (3k images, ablation) | FID: 31.30 -> 31.41 | No clear effect |
| add spatial swap to channel swap (ablation) | SSG · MS-COCO 2014 conditional (3k images, ablation) | ImageReward: 0.286 -> 0.297 | Improved |
| add CFG on top of SSG (ablation) | SSG · MS-COCO 2014 conditional (3k images, ablation) | ImageReward: 0.297 -> 0.492 | Improved |
Closest papers by meaning
- Token Perturbation Guidance for Diffusion Models (NeurIPS 2025 · similarity 0.94)
- CFG++: Manifold-constrained Classifier Free Guidance for Diffusion Models (ICLR 2025 · similarity 0.93)
- Eliminating Oversaturation and Artifacts of High Guidance Scales in Diffusion Models (ICLR 2025 · similarity 0.93)
- No Training, No Problem: Rethinking Classifier-Free Guidance for Diffusion Models (ICLR 2025 · similarity 0.93)
- Adaptive Guidance: Training-free Acceleration of Conditional Diffusion Models (AAAI 2025 · similarity 0.92)
- Elucidating the design space of classifier-guided diffusion generation (ICLR 2024 · similarity 0.92)
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.
