Research map / Image generation / Text-to-image diffusion
i-Design: Step-by-Step Graphic Layout Design with Progressive Aesthetic Policy Optimization
Sohan Patnaik, Rishabh Jain, Balaji Krishnamurthy, Mausoom Sarkar
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 PAPO after PIL (ablation) | i-Design-8B · Crello | Mean IoU: 44.39 -> 48.27 | Improved |
| add PAPO after PIL (ablation) | i-Design-8B · Crello | Win rate (GPT-4o judge, %): 20.41 -> 27.64 | Improved |
| add PAPO after PIL (ablation) | i-Design-8B · Crello | Win rate (o3 judge, %): 15.28 -> 19.03 | Improved |
| use PAPO instead of AAPA as the RL stage (ablation) | i-Design-8B · Crello | Mean IoU: 46.62 -> 48.27 | Improved |
| add ground-truth anchor to preference graph (ablation) | i-Design-8B · Crello | Mean IoU: 45.19 -> 48.27 | Improved |
| add ground-truth anchor to preference graph (ablation) | i-Design-8B · Crello | Win rate (GPT-4o judge, %): 23.04 -> 27.64 | Improved |
| progressive placement with k=sqrt(N) instead of single pass k=N (ablation) | i-Design-8B · Crello | Mean IoU: 40.28 -> 48.27 | Improved |
| progressive placement with k=sqrt(N) instead of single pass k=N (ablation) | i-Design-8B · Crello | Win rate (GPT-4o judge, %): 17.64 -> 27.64 | Improved |
| use k=1 instead of k=sqrt(N) (ablation) | i-Design-8B · Crello | Mean IoU: 48.27 -> 47.72 | No clear effect |
| z-order grouping instead of random grouping (ablation) | i-Design-8B · Crello | Mean IoU: 43.27 -> 48.27 | Improved |
| z-order grouping instead of random grouping (ablation) | i-Design-8B · Crello | Win rate (o3 judge, %): 6.11 -> 19.03 | Improved |
| semantic GPT-4o grouping instead of z-order (ablation) | i-Design-8B · Crello | Win rate (GPT-4o judge, %): 27.64 -> 29.23 | Mixed |
Closest papers by meaning
- AesthetiQ: Enhancing Graphic Layout Design via Aesthetic-Aware Preference Alignment of Multi-modal Large Language Models (CVPR 2025 · similarity 0.89)
- Towards Aligned Layout Generation via Diffusion Model with Aesthetic Constraints (ICLR 2024 · similarity 0.88)
- PrefPaint: Aligning Image Inpainting Diffusion Model with Human Preference (NeurIPS 2024 · similarity 0.87)
- Graphic Design with Large Multimodal Model (AAAI 2025 · similarity 0.87)
- LineArt: A Knowledge-guided Training-free High-quality Appearance Transfer for Design Drawing with Diffusion Model (CVPR 2025 · similarity 0.86)
- Seeing is Improving: Visual Feedback for Iterative Text Layout Refinement (CVPR 2026 · 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.
