Coarse-to-Fine Compositional Diffusion for Long-Horizon Planning
Byoungwoo Park, Utkarsh A. Mishra, Jaemoo Choi, Juho Lee, Yongxin Chen · 2026
Read Coarse-to-Fine Compositional Diffusion for Long-Horizon Planning by Byoungwoo Park, Utkarsh A. Mishra, Jaemoo Choi, Juho Lee, Yongxin Chen (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-00837.txt.
Description
Park and co-authors plan over long horizons with diffusion models in this 2026 paper. Diffusion models give strong priors for structured output. But many tasks need results far past the scale the models trained on. A common fix composes short local plans into one long output. Joined plans can agree with their neighbors yet form a poor whole.
Their method, CoFi, splits the work by scale at run time. First it aligns local drafts around a shared coarse scaffold. That scaffold holds the long-range shape of the route or task chain. CoFi then adds noise back to a mid level and denoises with the same local prior. Fine detail returns while the global shape holds. Older fixes repeat local passes or optimize at run time, which grows costly fast. Tests span long-horizon robot planning and more. The authors present CoFi as a cheaper path to coherent plans.
Catalog record
| CREATOR | Byoungwoo Park, Utkarsh A. Mishra, Jaemoo Choi, Juho Lee, Yongxin Chen |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.00837