Research map / Image generation / Fast diffusion sampling
Visual Diffusion Models are Geometric Solvers
Nir Goren, Shai Yehezkel, Omer Dahary, Andrey Voynov, Or Patashnik, Daniel Cohen-Or
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 rigid snapping post-processing to predicted squares (ablation) | Visual diffusion for Inscribed Square · Inscribed Square (synthetic curves) | alignment A(S,C): -1.60 -> -0.90 | Improved |
| add rigid snapping post-processing to predicted squares (ablation) | Visual diffusion for Inscribed Square · Inscribed Square (synthetic curves) | squareness Q: 0.892 -> 0.891 | No clear effect |
| test on more input points than seen in training (10-20 -> 41-50) | Visual diffusion for Steiner Tree · Steiner Tree | valid rate: 0.996 -> 0.334 | Hurt |
| test on more input points than seen in training (10-20 -> 41-50) | Visual diffusion for Steiner Tree · Steiner Tree | length ratio to optimal (valid solutions): 1.0008 -> 1.0092 | Mixed |
| compare diffusion against MST baseline on Steiner tree length | Visual diffusion for Steiner Tree · Steiner Tree (10-20 points) | length ratio to optimal: 1.036 -> 1.0008 | Improved |
| test on 13-15 points instead of 7-12 (beyond training range) | Visual diffusion for Max Area Polygon · Max Area Polygon | valid rate: 0.953 -> 0.620 | Hurt |
| test on 13-15 points instead of 7-12 (beyond training range) | Visual diffusion for Max Area Polygon · Max Area Polygon | optimal solution rate: 0.574 -> 0.062 | Hurt |
| use diffusion instead of random simple polygons | Max Area Polygon · Max Area Polygon (7-12 points) | area ratio to optimal: 0.771 -> 0.988 | Improved |
Closest papers by meaning
- GeoX: Geometric Problem Solving Through Unified Formalized Vision-Language Pre-training (ICLR 2025 · similarity 0.86)
- Hyperbolic Geometric Latent Diffusion Model for Graph Generation (ICML 2024 · similarity 0.86)
- Geometry Image Diffusion: Fast and Data-Efficient Text-to-3D with Image-Based Surface Representation (ICLR 2025 · similarity 0.85)
- SweetDreamer: Aligning Geometric Priors in 2D diffusion for Consistent Text-to-3D (ICLR 2024 · similarity 0.84)
- Geometric Trajectory Diffusion Models (NeurIPS 2024 · similarity 0.84)
- VisioMath: Benchmarking Figure-based Mathematical Reasoning in LMMs (ICLR 2026 · similarity 0.84)
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.
