Research map / Image generation / Autoregressive and latent generation
Dolfin: Diffusion Layout Transformers without Autoencoder
Yilin Wang, Zeyuan Chen, Liangjun Zhong, Zheng Ding, Zhuowen Tu
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 |
|---|---|---|---|
| operate directly on input space instead of adding an MLP autoencoder latent space (ablation) | Dolfin · PublayNet unconditional | FID: 15.50 -> 9.12 | Improved |
| operate directly on input space instead of MLP autoencoder (ablation) | Dolfin · PublayNet unconditional | Alignment: 0.129 -> 0.074 | Improved |
| increase transformer size from 4x512 to 8x512 (ablation) | Dolfin · PublayNet unconditional | FID: 9.12 -> 9.45 | No clear effect |
| reduce hidden size from 512 to 384 (ablation) | Dolfin · PublayNet unconditional | FID: 9.12 -> 9.49 | No clear effect |
| use autoregressive diffusion (Dolfin-AR) instead of non-autoregressive | Dolfin · PublayNet unconditional | Alignment: 0.074 -> 0.042 | Improved |
| use autoregressive diffusion (Dolfin-AR) instead of non-autoregressive | Dolfin · PublayNet unconditional | FID: 9.12 -> 16.88 | Hurt |
| use Dolfin-AR on RICO unconditional | Dolfin · RICO unconditional | MaxIoU: 0.42 -> 0.17 | Hurt |
| apply Dolfin to heavily overlapping RICO layouts | Dolfin · RICO unconditional | FID: 13.00 (PLay) vs 15.38 | Hurt |
| Dolfin versus LT and LayoutGAN++ on line segment generation | line segments · ShanghaiTech Wireframe | FID: 317.4 -> 92.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 2 on most shared results.
- PLay — PLay: Parametrically Conditioned Layout Generation using Latent Diffusion (ICML 2023) · better on 1 of 2 shared results
- MaskGIT — MaskGIT: Masked Generative Image Transformer (CVPR 2022) · better on 0 of 1 shared result
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
- Hybrid Layout Control for Diffusion Transformer: Fewer Annotations, Superior Aesthetics (ICCV 2025 · similarity 0.89)
- Towards Aligned Layout Generation via Diffusion Model with Aesthetic Constraints (ICLR 2024 · similarity 0.87)
- Towards Transformer-Based Aligned Generation with Self-Coherence Guidance (CVPR 2025 · similarity 0.85)
- Visual Layout Composer: Image-Vector Dual Diffusion Model for Design Layout Generation (CVPR 2024 · similarity 0.85)
- Adversarial Supervision Makes Layout-to-Image Diffusion Models Thrive (ICLR 2024 · similarity 0.85)
- MaskPLAN: Masked Generative Layout Planning from Partial Input (CVPR 2024 · similarity 0.85)
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.
