CMAD: Cooperative Multi-Agent Diffusion via Stochastic Optimal Control
This addresses a problem in generative modeling for researchers and practitioners by enabling more flexible composition of pre-trained models, though it appears incremental as it builds on existing diffusion and control frameworks.
The paper tackles the challenge of controlling the composition of multiple pre-trained continuous-time generative models by formulating it as a cooperative stochastic optimal control problem, where diffusion models are treated as interacting agents steered toward a shared objective, and validates this approach on conditional MNIST generation with comparisons to a baseline.
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 naive inference-time DPS-style baseline replacing learned cooperative control with per-step gradient guidance.