Research map / Image generation / Fast diffusion sampling
Extendable Planning via Multiscale Diffusion
Chang Chen, Hany Hamed, Doojin Baek, Taegu Kang, Samyeul Noh, Yoshua Bengio, Sungjin Ahn
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 |
|---|---|---|---|
| train Diffuser on PTE-extended data | Diffuser · Maze2D Large (single-task) | normalized score: 45.8 -> 114.4 | Improved |
| train Diffuser on PTE-extended data | Diffuser · Maze2D XXLarge (single-task) | normalized score: 14.1 -> 23.0 | Improved |
| use hierarchical planning (HD) instead of flat Diffuser | Diffuser-X · Maze2D XXLarge (single-task) | normalized score: 23.0 -> 56.4 | Improved |
| add multiscale levels, adaptive pondering and recursive sharing to HD | HD-X · Maze2D XXLarge (single-task) | normalized score: 56.4 -> 82.1 | Improved |
| add multiscale levels, adaptive pondering and recursive sharing to HD | HD-X · Multi2D Giant (multi-task) | normalized score: 168.8 -> 246.7 | Improved |
| train Diffuser on PTE-extended data | Diffuser · Extendable Franka Kitchen | average score: 43.8 -> 45.8 | Improved |
| add multiscale planning to HD | HD-X · Extendable Franka Kitchen | average score: 55.0 -> 59.2 | Improved |
| train Diffuser on PTE-extended data | Diffuser · Extendable Hopper | average score: 41.9 -> 51.2 | Improved |
| train Diffuser on PTE-extended data | Diffuser · Extendable Walker2d | average score: 54.4 -> 54.3 | No clear effect |
| use exponential instead of linear PTE (ablation) | HM-Diffuser · Maze2D-XXLarge, 7 PTE rounds | normalized score: 82.1 -> 30.0 (1M); 89.9 -> 120.5 (4M) | Mixed |
| increase data per round from 1M to 4M (ablation) | Linear PTE · Maze2D-XXLarge, HM-Diffuser | normalized score: 82.1 -> 89.9 | Improved |
| add more PTE rounds (round 1 to round 3) (ablation) | Diffuser-X · Extendable Franka Kitchen average | average score: 45.8 -> 31.7 -> 15.8 | Hurt |
Closest papers by meaning
- Compositional Monte Carlo Tree Diffusion for Extendable Planning (NeurIPS 2025 · similarity 0.89)
- DiffuserLite: Towards Real-time Diffusion Planning (NeurIPS 2024 · similarity 0.89)
- Simple Hierarchical Planning with Diffusion (ICLR 2024 · similarity 0.89)
- State-Covering Trajectory Stitching for Diffusion Planners (NeurIPS 2025 · similarity 0.89)
- Generative Trajectory Stitching through Diffusion Composition (NeurIPS 2025 · similarity 0.89)
- Coarse-to-Fine Compositional Diffusion for Long-Horizon Planning (NeurIPS 2026 · similarity 0.89)
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.
