Efficient Generative Modeling beyond Memoryless Diffusion via Adjoint Schrödinger Bridge Matching

Jeongwoo Shin, Jinhwan Sul, Joonseok Lee, Jaewoong Choi, Jaemoo Choi
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:112652-112668, 2026.

Abstract

Diffusion models often yield highly curved trajectories and noisy score targets due to an uninformative, memoryless forward process that induces independent data-noise coupling. We propose Adjoint Schrödinger Bridge Matching (ASBM), a generative modeling framework that recovers optimal trajectories in high dimensions via two stages. First, we view the Schrödinger Bridge (SB) forward dynamic as a coupling construction problem and learn it through a data-to-energy sampling perspective that transports data to an energy-defined prior. Then, we learn the backward generative dynamic with a simple matching loss supervised by the induced optimal coupling. By operating in a non-memoryless regime, ASBM produces significantly straighter and more efficient sampling paths. Compared to prior works, ASBM scales to high-dimensional data with notably improved stability and efficiency. Extensive experiments on image generation show that ASBM improves fidelity with fewer sampling steps. We further showcase the effectiveness of our optimal trajectory via distillation to a one-step generator.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-shin26d, title = {Efficient Generative Modeling beyond Memoryless Diffusion via Adjoint Schrödinger Bridge Matching}, author = {Shin, Jeongwoo and Sul, Jinhwan and Lee, Joonseok and Choi, Jaewoong and Choi, Jaemoo}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {112652--112668}, 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/shin26d/shin26d.pdf}, url = {https://proceedings.mlr.press/v306/shin26d.html}, abstract = {Diffusion models often yield highly curved trajectories and noisy score targets due to an uninformative, memoryless forward process that induces independent data-noise coupling. We propose Adjoint Schrödinger Bridge Matching (ASBM), a generative modeling framework that recovers optimal trajectories in high dimensions via two stages. First, we view the Schrödinger Bridge (SB) forward dynamic as a coupling construction problem and learn it through a data-to-energy sampling perspective that transports data to an energy-defined prior. Then, we learn the backward generative dynamic with a simple matching loss supervised by the induced optimal coupling. By operating in a non-memoryless regime, ASBM produces significantly straighter and more efficient sampling paths. Compared to prior works, ASBM scales to high-dimensional data with notably improved stability and efficiency. Extensive experiments on image generation show that ASBM improves fidelity with fewer sampling steps. We further showcase the effectiveness of our optimal trajectory via distillation to a one-step generator.} }
Endnote
%0 Conference Paper %T Efficient Generative Modeling beyond Memoryless Diffusion via Adjoint Schrödinger Bridge Matching %A Jeongwoo Shin %A Jinhwan Sul %A Joonseok Lee %A Jaewoong Choi %A Jaemoo Choi %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-shin26d %I PMLR %P 112652--112668 %U https://proceedings.mlr.press/v306/shin26d.html %V 306 %X Diffusion models often yield highly curved trajectories and noisy score targets due to an uninformative, memoryless forward process that induces independent data-noise coupling. We propose Adjoint Schrödinger Bridge Matching (ASBM), a generative modeling framework that recovers optimal trajectories in high dimensions via two stages. First, we view the Schrödinger Bridge (SB) forward dynamic as a coupling construction problem and learn it through a data-to-energy sampling perspective that transports data to an energy-defined prior. Then, we learn the backward generative dynamic with a simple matching loss supervised by the induced optimal coupling. By operating in a non-memoryless regime, ASBM produces significantly straighter and more efficient sampling paths. Compared to prior works, ASBM scales to high-dimensional data with notably improved stability and efficiency. Extensive experiments on image generation show that ASBM improves fidelity with fewer sampling steps. We further showcase the effectiveness of our optimal trajectory via distillation to a one-step generator.
APA
Shin, J., Sul, J., Lee, J., Choi, J. & Choi, J.. (2026). Efficient Generative Modeling beyond Memoryless Diffusion via Adjoint Schrödinger Bridge Matching. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:112652-112668 Available from https://proceedings.mlr.press/v306/shin26d.html.

Related Material