Abstract:Continuous-time generative models have achieved remarkable success in image restoration and synthesis. However, controlling the composition of multiple pre-trained models remains an open challenge. Current approaches largely treat composition as an algebraic composition of probability densities, such as via products or mixtures of experts. This perspective assumes the target distribution is known explicitly, which is almost never the case. In this work, we propose a different paradigm that formulates compositional generation as a cooperative Stochastic Optimal Control problem. Rather than combining probability densities, we treat pre-trained diffusion models as interacting agents whose diffusion trajectories are jointly steered, via optimal control, toward a shared objective defined on their aggregated output. We validate our framework on conditional MNIST generation and compare it against a naïve inference-time DPS-style baseline replacing learned cooperative control with per-step gradient guidance.
| Subjects: | Machine Learning (cs.LG) |
| MSC classes: | 93E20 |
| Cite as: | arXiv:2602.10933 [cs.LG] |
| (or arXiv:2602.10933v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2602.10933 arXiv-issued DOI via DataCite |
From: Alexander Denker [view email]
[v1]
Wed, 11 Feb 2026 15:12:43 UTC (1,743 KB)
[v2]
Tue, 19 May 2026 13:50:37 UTC (1,744 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。