Research map / Image generation / Autoregressive and latent generation
RealUHR: Harnessing Patch-Cascade Flows for Photorealistic Ultra-High-Resolution Synthesis
Yongsheng Yu, Haitian Zheng, Zhe Lin, Connelly Barnes, Yuqian Zhou, Zhifei Zhang, Jiebo Luo
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 non-uniform time schedule vs uniform (ablation) | RealUHR · UHR benchmark 2K | Patch FID: 29.72 -> 26.09 | Improved |
| add non-uniform time schedule vs cosine (ablation) | RealUHR · UHR benchmark 2K | QualiCLIP: 0.492 -> 0.495 | No clear effect |
| add Patch Initialization from anchor patches (ablation) | RealUHR · UHR benchmark 4K | FID: 145.82 -> 139.29 | Improved |
| add Patch Initialization from anchor patches (ablation) | RealUHR · UHR benchmark 4K | Patch FID: 53.11 -> 46.15 | Improved |
| add Guidance-Consistent Adaptation (ablation) | RealUHR · UHR benchmark 4K | Patch FID: 50.48 -> 46.15 | Improved |
| add GCA Stage 1 reverse distillation (ablation) | RealUHR · UHR benchmark 4K | FID: 141.03 -> 139.29 | Improved |
| add GCA Stage 1 reverse distillation (ablation) | RealUHR · UHR benchmark 2K | Patch FID: 29.03 -> 26.10 | Improved |
| add Guidance-Consistent Adaptation (ablation) | RealUHR · UHR benchmark 2K | FID: 42.14 -> 40.29 | Improved |
| patch-cascade pipeline vs I-Max on same FLUX backbone | RealUHR · UHR benchmark 4K | Time (s): 751 -> 108 | Improved |
| RealUHR vs UltraPixel on CLIPScore | RealUHR · UHR benchmark 4K | CLIPScore: 34.83 -> 34.19 | Hurt |
| RealUHR vs Diffusion-4K on 2K time | RealUHR · UHR benchmark 2K | Time (s): 22 -> 38 | 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.
- UltraPixel — UltraPixel: Advancing Ultra High-Resolution Image Synthesis to New Peaks (NeurIPS 2024) · 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
- EditBridge: Towards Faithful and Efficient Ultra-High-Resolution Image Editing (NeurIPS 2026 · similarity 0.88)
- UltraPixel: Advancing Ultra High-Resolution Image Synthesis to New Peaks (NeurIPS 2024 · similarity 0.87)
- Ultra-Resolution Adaptation with Ease (ICML 2025 · similarity 0.87)
- UltraHR-100K: Enhancing UHR Image Synthesis with A Large-Scale High-Quality Dataset (NeurIPS 2025 · similarity 0.87)
- HiFlow: Training-free High-Resolution Image Generation with Flow-Aligned Guidance (NeurIPS 2025 · similarity 0.87)
- PatchVSR: Breaking Video Diffusion Resolution Limits with Patch-wise Video Super-Resolution (CVPR 2025 · similarity 0.86)
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.
