Coupled Inference in Diffusion Models for Semantic Decomposition
Calvin Yeung, Ali Zakeri, Zhuowen Zou, Mohsen ImaniDepartment of Computer ScienceUniversity of California, IrvineIrvine, CA 92697 · 2026
Read Coupled Inference in Diffusion Models for Semantic Decomposition by Calvin Yeung, Ali Zakeri, Zhuowen Zou, Mohsen ImaniDepartment of Computer ScienceUniversity of California, IrvineIrvine, CA 92697 (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-2602-09983.txt.
Description
Scenes are built from parts, and understanding a scene often means pulling those parts back apart. A 2026 paper led by Calvin Yeung, with three co-authors, tackles that decomposition problem inside diffusion models. The team treats decomposition as an inverse problem, coupling diffusion processes with a guidance term that pushes factor estimates to rebuild the bound vector.
A new iterative sampling scheme lifts performance further. The authors also show that attention-based resonator networks fall out as a special case of their framework. Across synthetic semantic decomposition tasks, the coupled inference method outperforms resonator networks, linking Hopfield-style decomposition to modern diffusion machinery.
Catalog record
| CREATOR | Calvin Yeung, Ali Zakeri, Zhuowen Zou, Mohsen ImaniDepartment of Computer ScienceUniversity of California, IrvineIrvine, CA 92697 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-09983