Research map / Image generation / Autoregressive and latent generation
LookFlow: Training-Free and Efficient High-Resolution Image Synthesis via Dynamic Lookahead Guidance Flow
Yuan Zhou, Yan Zhang, Jianlong Chang, Xin Gu, Ying Wang, Kun Ding, Guangwen Yang, Shiming Xiang
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 |
|---|---|---|---|
| use lookahead step s=2 instead of s=1 (adjacent-step guidance) (ablation) | LookFlow DLG · COCO 2048x2048 | FID: 61.138 -> 60.815 | Improved |
| increase lookahead step from 2 to 5 (ablation) | LookFlow DLG · COCO 2048x2048 | FID: 60.815 -> 64.207 | Hurt |
| add dynamic lookahead guidance (DLG) (ablation) | LookFlow · COCO | image quality | Improved |
| add sparse recomputation and caching (SRC) (ablation) | LookFlow · COCO | FID / inference time: up to 2.01x speedup | Improved |
| cache attention and MLP features (Attn & MLP caching) (ablation) | Flux · COCO | FID: 60.39 -> 60.82 | No clear effect |
| increase cache stride from 4 to 5 (ablation) | LookFlow SRC · COCO 2048x2048 | FID: 60.815 -> 60.599 | No clear effect |
| use cache stride 4 versus 2 (ablation) | LookFlow SRC · COCO 2048x2048 | FID: 61.189 -> 60.815 | No clear effect |
| use LookFlow instead of I-Max | FLUX.1-dev at 4096x4096 · COCO 4096x4096 | FID: 65.924 -> 59.433 | Improved |
| use LookFlow instead of I-Max | FLUX.1-dev at 4096x4096 · COCO 4096x4096 | Time (s): 435 -> 222 | Improved |
| use LookFlow instead of Flux direct inference | FLUX.1-dev at 2048x2048 · COCO 2048x2048 | FID: 59.204 -> 60.815 | Hurt |
| use LookFlow instead of Flux direct inference | FLUX.1-dev at 3072x3072 · COCO 3072x3072 | FID: 203.495 -> 59.597 | Improved |
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.
- DemoFusion — DemoFusion: Democratising High-Resolution Image Generation With No $$$ (CVPR 2024) · better on 3 of 3 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
- HiFlow: Training-free High-Resolution Image Generation with Flow-Aligned Guidance (NeurIPS 2025 · similarity 0.90)
- FMBoost: Boosting Latent Diffusion with Flow Matching (ECCV 2024 · similarity 0.88)
- Scaling Rectified Flow Transformers for High-Resolution Image Synthesis (ICML 2024 · similarity 0.88)
- Boosting Flow-based Generative Super-Resolution Models via Learned Prior (CVPR 2024 · similarity 0.88)
- STARFlow: Scaling Latent Normalizing Flows for High-resolution Image Synthesis (NeurIPS 2025 · similarity 0.87)
- Native-Resolution Image Synthesis (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.
