Research map / Image generation / Fast diffusion sampling
Improving Diffusion Generalization with Weak-to-Strong Segmented Guidance
Liangyu Yuan, Yufei Huang, Mingkun Lei, Tong Zhao, Ruoyu Wang, Changxi Chi, Yiwei Wang, Chi Zhang
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 |
|---|---|---|---|
| segmented guidance (CFG then SLG) vs CFG alone (ablation) | SD3.5-medium · MS-COCO-1K | Aesthetic score: 5.279 -> 5.717 | Improved |
| segmented guidance vs CFG alone | SD3.5-medium · MS-COCO-1K | HPSv2.1: 29.199 -> 29.736 | Improved |
| segmented guidance vs SLG alone | SD3.5-medium · MS-COCO-1K | HPSv2.1: 27.295 -> 29.736 | Improved |
| SGG vs S2 Guidance | SD3-medium · MS-COCO-1K | HPSv2.1: 29.614 -> 29.541 | No clear effect |
| add weak-to-strong guidance to training target (SGG) (ablation) | SiT-B/2 · ImageNet 256 conditional | FID: 31.22 -> 4.58 | Improved |
| use SGG (CFG+BR segmented) instead of MG-only training target (ablation) | SiT-B/2 · ImageNet 256 conditional | FID: 5.88 -> 4.58 | Improved |
| add BR branch weak signal in training (ablation) | SiT-B/2 · ImageNet 256 unconditional | FID: 61.27 -> 43.25 | Improved |
| add AG weak model in training (ablation) | SiT-B/2 · ImageNet 256 unconditional | FID: 61.27 -> 45.97 | Improved |
| add SGG on top of REPA (ablation) | SiT-B/2+REPA · ImageNet 256 conditional | FID: 20.46 -> 3.07 | Improved |
| move segmentation timestep tau from 0 to 0.2 (ablation) | SGG training · ImageNet 256 conditional | FID: 5.88 -> 4.58 | Improved |
| move segmentation timestep tau from 0.2 to 0.4 (ablation) | SGG training · ImageNet 256 conditional | FID: 4.58 -> 5.79 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- S2 Guidance — S$^2$-Guidance: Stochastic Self-Guidance for Training-Free Enhancement of Diffusion Models (ICLR 2026) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Eliminating Oversaturation and Artifacts of High Guidance Scales in Diffusion Models (ICLR 2025 · similarity 0.92)
- Rectified Diffusion Guidance for Conditional Generation (CVPR 2025 · similarity 0.91)
- Diffusion Models without Classifier-free Guidance (NeurIPS 2026 · similarity 0.91)
- Adaptive Guidance: Training-free Acceleration of Conditional Diffusion Models (AAAI 2025 · similarity 0.91)
- Feedback Guidance of Diffusion Models (NeurIPS 2025 · similarity 0.91)
- Understanding and Improving Training-free Loss-based Diffusion Guidance (NeurIPS 2024 · similarity 0.90)
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.
