Latent-Guided Cooperative Energy-Based Models

Cong Geng, Xue Han, Ye Yuan, Qiang Hu, Xin Huang, Ruiqiao Bai, Junlan Feng
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:34544-34566, 2026.

Abstract

Energy-based models (EBMs) provide a flexible framework for generative models with strong distribution modeling capabilities. Nevertheless, their broader adoption has been limited by the difficulty of stable and efficient training. In this paper, we propose a unified and efficient latent-guided cooperative EBM that leverages informative target latent variables to guide the joint energy in capturing both data distribution and semantic structure, along with a cooperative generator designed for effective MCMC initialization. Our joint space optimization only requires MCMC sampling in the data space, and allows the energy to learn semantic data–latent relationships directly from real data. Experiments show our method improves generation quality and training stability with fewer resources, and performs effectively across multiple downstream tasks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-geng26b, title = {Latent-Guided Cooperative Energy-Based Models}, author = {Geng, Cong and Han, Xue and Yuan, Ye and Hu, Qiang and Huang, Xin and Bai, Ruiqiao and Feng, Junlan}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {34544--34566}, 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/geng26b/geng26b.pdf}, url = {https://proceedings.mlr.press/v306/geng26b.html}, abstract = {Energy-based models (EBMs) provide a flexible framework for generative models with strong distribution modeling capabilities. Nevertheless, their broader adoption has been limited by the difficulty of stable and efficient training. In this paper, we propose a unified and efficient latent-guided cooperative EBM that leverages informative target latent variables to guide the joint energy in capturing both data distribution and semantic structure, along with a cooperative generator designed for effective MCMC initialization. Our joint space optimization only requires MCMC sampling in the data space, and allows the energy to learn semantic data–latent relationships directly from real data. Experiments show our method improves generation quality and training stability with fewer resources, and performs effectively across multiple downstream tasks.} }
Endnote
%0 Conference Paper %T Latent-Guided Cooperative Energy-Based Models %A Cong Geng %A Xue Han %A Ye Yuan %A Qiang Hu %A Xin Huang %A Ruiqiao Bai %A Junlan Feng %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-geng26b %I PMLR %P 34544--34566 %U https://proceedings.mlr.press/v306/geng26b.html %V 306 %X Energy-based models (EBMs) provide a flexible framework for generative models with strong distribution modeling capabilities. Nevertheless, their broader adoption has been limited by the difficulty of stable and efficient training. In this paper, we propose a unified and efficient latent-guided cooperative EBM that leverages informative target latent variables to guide the joint energy in capturing both data distribution and semantic structure, along with a cooperative generator designed for effective MCMC initialization. Our joint space optimization only requires MCMC sampling in the data space, and allows the energy to learn semantic data–latent relationships directly from real data. Experiments show our method improves generation quality and training stability with fewer resources, and performs effectively across multiple downstream tasks.
APA
Geng, C., Han, X., Yuan, Y., Hu, Q., Huang, X., Bai, R. & Feng, J.. (2026). Latent-Guided Cooperative Energy-Based Models. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:34544-34566 Available from https://proceedings.mlr.press/v306/geng26b.html.

Related Material