Research map / 3D vision / 3D generation
WorldGrow: Generating Infinite 3D World
Sikuang Li, Chen Yang, Jiemin Fang, Taoran Yi, Jia Lu, Jiazhong Cen, Lingxi Xie, Wei Shen, Qi Tian
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 occlusion-aware aggregation without retraining decoder (ablation) | TRELLIS SLAT VAE · scene block reconstruction | PSNR: 23.17 -> 22.23 | Hurt |
| retrain the decoder on scene blocks (ablation) | TRELLIS SLAT VAE · scene block reconstruction | PSNR: 23.17 -> 25.84 | Improved |
| combine occlusion-aware aggregation with retrained decoder (ablation) | TRELLIS SLAT VAE · scene block reconstruction | PSNR: 23.17 -> 31.32 | Improved |
| add data curation (filter scene blocks) (ablation) | WorldGrow · 3D-FRONT scene blocks | FID: 9.09 -> 7.52 | Improved |
| add coarse scene generation (coarse-to-fine) (ablation) | WorldGrow · 3D-FRONT scene blocks | FID: 17.04 -> 7.52 | Improved |
| add coarse scene generation (ablation) | WorldGrow · 3D-FRONT scene blocks | MMD-EMD: 13.62 -> 13.33 | No clear effect |
| fine-tune TRELLIS on 3D-FRONT | TRELLIS · 3D-FRONT scene blocks | FID: 53.49 -> 24.61 | Improved |
| add block inpainting and scene-friendly SLAT to fine-tuned TRELLIS | TRELLIS fine-tuned · 3D-FRONT scene blocks | FID: 24.61 -> 7.52 | Improved |
| use WorldGrow instead of SynCity | infinite scene generation · generation speed | time per block: 2 minutes -> 20 seconds | Improved |
| use WorldGrow instead of SynCity | textured indoor scenes · human preference | structure plausibility (1-5): 2.48 -> 4.48 | 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.
- SynCity — SynCity: Training-Free Generation of 3D Worlds (ICCV 2025) · better on 6 of 6 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
- WorldFlow3D: Flowing Through 3D Distributions for Unbounded World Generation (ECCV 2026 · similarity 0.91)
- WorldMesh: Generating Navigable Multi-Room 3D Scenes via Mesh-Conditioned Image Diffusion (ECCV 2026 · similarity 0.90)
- WonderWorld: Interactive 3D Scene Generation from a Single Image (CVPR 2025 · similarity 0.89)
- A 3D Scene is Worth 1K Tokens: 3D-Grounded Representation for Scene Generation at Scale (NeurIPS 2026 · similarity 0.88)
- WorldCoder-Bench: Benchmarking Physically Grounded 3D World Synthesis (NeurIPS 2026 · similarity 0.88)
- 3D-SceneDreamer: Text-Driven 3D-Consistent Scene Generation (CVPR 2024 · similarity 0.88)
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.
