Research map / Image generation / Autoregressive and latent generation
Puzzlefusion: Unleashing the Power of Diffusion Models for Spatial Puzzle Solving
Sepidehsadat Hosseini, Mohammad Amin Shabani, Saghar Irandoust, Yasutaka Furukawa
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 |
|---|---|---|---|
| remove G-SA (ablation) | PuzzleFusion · Full MagicPlan RLA | MPE: 56.9 -> 40.8 | Improved |
| add P-SA (ablation) | PuzzleFusion · Full MagicPlan RLA | MPE: 45.2 -> 40.8 | Improved |
| add door matching loss (doors) (ablation) | PuzzleFusion · Full MagicPlan RLA | MPE: 43.4 -> 40.8 | Improved |
| matching loss on doors vs all corners (ablation) | PuzzleFusion · Full MagicPlan RLA | GED: 3.6 -> 3.1 | Mixed |
| redundant per-corner representation vs compact per-room node (ablation) | PuzzleFusion · Full MagicPlan RLA | MPE: 51.62 -> 41.23 | Improved |
| remove room type at test time (ablation) | PuzzleFusion · Full MagicPlan RLA | MPE: 40.8 -> 48.4 | Hurt |
| remove door information at test time (ablation) | PuzzleFusion · Full MagicPlan RLA | GED: 3.1 -> 5.2 | Hurt |
| noisy room types at test time (ablation) | PuzzleFusion · Full MagicPlan RLA | MPE: 40.8 -> 49.7 | Hurt |
| add duplicated/enlarged rooms to data (ablation) | PuzzleFusion · Full MagicPlan RLA | MPE: 40.8 -> 48.3 | Hurt |
| test-time noise level 0 to 2 (trained noise-free) | PuzzleFusion vs Harel et al. · Cross-cut Jigsaw Puzzle | Overlap: 0.94 -> 0.93 vs 0.91 -> 0.30 | Improved |
| add door matching loss (ablation) | PuzzleFusion · Voronoi Jigsaw Puzzle | Overlap: 0.65 -> 0.70 | Improved |
| add door matching loss (ablation) | PuzzleFusion · Cross-cut Jigsaw Puzzle | Recall: 0.77 -> 0.91 | Improved |
Closest papers by meaning
- Solving Masked Jigsaw Puzzles with Diffusion Vision Transformers (CVPR 2024 · similarity 0.85)
- OptiScene: LLM-driven Indoor Scene Layout Generation via Scaled Human-aligned Data Synthesis and Multi-Stage Preference Optimization (NeurIPS 2025 · similarity 0.84)
- ERL-MPP: Evolutionary Reinforcement Learning with Multi-head Puzzle Perception for Solving Large-scale Jigsaw Puzzles of Eroded Gaps (AAAI 2025 · similarity 0.84)
- PuzzleFusion++: Auto-agglomerative 3D Fracture Assembly by Denoise and Verify (ICLR 2025 · similarity 0.84)
- Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoning (NeurIPS 2025 · similarity 0.83)
- PuzzleWorld: A Benchmark for Multimodal, Open-Ended Reasoning in Puzzlehunts (ICLR 2026 · similarity 0.83)
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.
