Diffusion Controller: Framework, Algorithms and Parameterization

Tong Yang, Moonkyung Ryu, Chih-Wei Hsu, Guy Tennenholtz, Yuejie Chi, Craig Boutilier, Bo Dai
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:145823-145854, 2026.

Abstract

Controllable diffusion generation often relies on various heuristics that are seemingly disconnected without a unified understanding. We bridge this gap with Diffusion Controller (DiffCon), a unified control-theoretic view that casts reverse diffusion sampling as state-only stochastic control within (generalized) linearly-solvable Markov Decision Processes (LS-MDPs). Under this framework, control acts by reweighting the pretrained reverse-time transition kernels, balancing terminal objectives against an $f$-divergence cost. From the resulting optimality conditions, we derive practical reinforcement learning methods for diffusion fine-tuning: (i) $f$-divergence-regularized policy-gradient updates, including a PPO-style rule, and (ii) a regularizer-determined reward-weighted regression objective with a minimizer-preservation guarantee under the Kullback–Leibler (KL) divergence. The LS-MDP framework further implies a principled model form: the optimal score decomposes into a fixed pretrained baseline plus a lightweight control correction, motivating a side-network parameterization conditioned on exposed intermediate denoising outputs, enabling effective gray-box adaptation with a frozen backbone. Experiments on Stable Diffusion v1.4 across supervised and reward-driven finetuning show consistent gains in preference-alignment win rates and improved quality–efficiency trade-offs versus gray-box baselines and even the parameter-efficient white-box adapter LoRA.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-yang26bk, title = {Diffusion Controller: Framework, Algorithms and Parameterization}, author = {Yang, Tong and Ryu, Moonkyung and Hsu, Chih-Wei and Tennenholtz, Guy and Chi, Yuejie and Boutilier, Craig and Dai, Bo}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {145823--145854}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/yang26bk/yang26bk.pdf}, url = {https://proceedings.mlr.press/v306/yang26bk.html}, abstract = {Controllable diffusion generation often relies on various heuristics that are seemingly disconnected without a unified understanding. We bridge this gap with Diffusion Controller (DiffCon), a unified control-theoretic view that casts reverse diffusion sampling as state-only stochastic control within (generalized) linearly-solvable Markov Decision Processes (LS-MDPs). Under this framework, control acts by reweighting the pretrained reverse-time transition kernels, balancing terminal objectives against an $f$-divergence cost. From the resulting optimality conditions, we derive practical reinforcement learning methods for diffusion fine-tuning: (i) $f$-divergence-regularized policy-gradient updates, including a PPO-style rule, and (ii) a regularizer-determined reward-weighted regression objective with a minimizer-preservation guarantee under the Kullback–Leibler (KL) divergence. The LS-MDP framework further implies a principled model form: the optimal score decomposes into a fixed pretrained baseline plus a lightweight control correction, motivating a side-network parameterization conditioned on exposed intermediate denoising outputs, enabling effective gray-box adaptation with a frozen backbone. Experiments on Stable Diffusion v1.4 across supervised and reward-driven finetuning show consistent gains in preference-alignment win rates and improved quality–efficiency trade-offs versus gray-box baselines and even the parameter-efficient white-box adapter LoRA.} }
Endnote
%0 Conference Paper %T Diffusion Controller: Framework, Algorithms and Parameterization %A Tong Yang %A Moonkyung Ryu %A Chih-Wei Hsu %A Guy Tennenholtz %A Yuejie Chi %A Craig Boutilier %A Bo Dai %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-yang26bk %I PMLR %P 145823--145854 %U https://proceedings.mlr.press/v306/yang26bk.html %V 306 %X Controllable diffusion generation often relies on various heuristics that are seemingly disconnected without a unified understanding. We bridge this gap with Diffusion Controller (DiffCon), a unified control-theoretic view that casts reverse diffusion sampling as state-only stochastic control within (generalized) linearly-solvable Markov Decision Processes (LS-MDPs). Under this framework, control acts by reweighting the pretrained reverse-time transition kernels, balancing terminal objectives against an $f$-divergence cost. From the resulting optimality conditions, we derive practical reinforcement learning methods for diffusion fine-tuning: (i) $f$-divergence-regularized policy-gradient updates, including a PPO-style rule, and (ii) a regularizer-determined reward-weighted regression objective with a minimizer-preservation guarantee under the Kullback–Leibler (KL) divergence. The LS-MDP framework further implies a principled model form: the optimal score decomposes into a fixed pretrained baseline plus a lightweight control correction, motivating a side-network parameterization conditioned on exposed intermediate denoising outputs, enabling effective gray-box adaptation with a frozen backbone. Experiments on Stable Diffusion v1.4 across supervised and reward-driven finetuning show consistent gains in preference-alignment win rates and improved quality–efficiency trade-offs versus gray-box baselines and even the parameter-efficient white-box adapter LoRA.
APA
Yang, T., Ryu, M., Hsu, C., Tennenholtz, G., Chi, Y., Boutilier, C. & Dai, B.. (2026). Diffusion Controller: Framework, Algorithms and Parameterization. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:145823-145854 Available from https://proceedings.mlr.press/v306/yang26bk.html.

Related Material