CoGenCast: A Coupled Autoregressive–Flow Generative Framework for Time Series Forecasting

Mingyue Cheng, Yaguo Liu, Daoyu Wang, Xiaoyu Tao, Qi Liu
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:18941-18965, 2026.

Abstract

Time series forecasting can be viewed as a generative problem that requires both semantic understanding over contextual conditions and stochastic modeling of continuous temporal dynamics. Existing approaches typically rely on either autoregressive large language models (LLMs) for semantic context modeling or diffusion-like models for continuous probabilistic generation. However, neither method alone can adequately model both aspects simultaneously. In this work, we propose CoGenCast, a hybrid generative framework that couples pre-trained LLMs with flow-matching mechanism for effective time series forecasting. Specifically, we reconfigure pre-trained decoder-only LLMs into a native forecasting encoder–decoder backbone by modifying only the attention topology, enabling bidirectional context encoding and causal representation generation. Building on this, a flow-matching mechanism is further integrated to model temporal evolution, capturing continuous stochastic dynamics conditioned on the autoregressively generated representation. Notably, CoGenCast naturally supports multimodal forecasting and cross-domain unified training. Extensive experiments on multiple benchmarks show that CoGenCast achieves competitive performance compared to previous baselines. Code is available at https://github.com/liuyaguo/_CoGenCast.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-cheng26h, title = {{C}o{G}en{C}ast: A Coupled Autoregressive–Flow Generative Framework for Time Series Forecasting}, author = {Cheng, Mingyue and Liu, Yaguo and Wang, Daoyu and Tao, Xiaoyu and Liu, Qi}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {18941--18965}, 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/cheng26h/cheng26h.pdf}, url = {https://proceedings.mlr.press/v306/cheng26h.html}, abstract = {Time series forecasting can be viewed as a generative problem that requires both semantic understanding over contextual conditions and stochastic modeling of continuous temporal dynamics. Existing approaches typically rely on either autoregressive large language models (LLMs) for semantic context modeling or diffusion-like models for continuous probabilistic generation. However, neither method alone can adequately model both aspects simultaneously. In this work, we propose CoGenCast, a hybrid generative framework that couples pre-trained LLMs with flow-matching mechanism for effective time series forecasting. Specifically, we reconfigure pre-trained decoder-only LLMs into a native forecasting encoder–decoder backbone by modifying only the attention topology, enabling bidirectional context encoding and causal representation generation. Building on this, a flow-matching mechanism is further integrated to model temporal evolution, capturing continuous stochastic dynamics conditioned on the autoregressively generated representation. Notably, CoGenCast naturally supports multimodal forecasting and cross-domain unified training. Extensive experiments on multiple benchmarks show that CoGenCast achieves competitive performance compared to previous baselines. Code is available at https://github.com/liuyaguo/_CoGenCast.} }
Endnote
%0 Conference Paper %T CoGenCast: A Coupled Autoregressive–Flow Generative Framework for Time Series Forecasting %A Mingyue Cheng %A Yaguo Liu %A Daoyu Wang %A Xiaoyu Tao %A Qi Liu %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-cheng26h %I PMLR %P 18941--18965 %U https://proceedings.mlr.press/v306/cheng26h.html %V 306 %X Time series forecasting can be viewed as a generative problem that requires both semantic understanding over contextual conditions and stochastic modeling of continuous temporal dynamics. Existing approaches typically rely on either autoregressive large language models (LLMs) for semantic context modeling or diffusion-like models for continuous probabilistic generation. However, neither method alone can adequately model both aspects simultaneously. In this work, we propose CoGenCast, a hybrid generative framework that couples pre-trained LLMs with flow-matching mechanism for effective time series forecasting. Specifically, we reconfigure pre-trained decoder-only LLMs into a native forecasting encoder–decoder backbone by modifying only the attention topology, enabling bidirectional context encoding and causal representation generation. Building on this, a flow-matching mechanism is further integrated to model temporal evolution, capturing continuous stochastic dynamics conditioned on the autoregressively generated representation. Notably, CoGenCast naturally supports multimodal forecasting and cross-domain unified training. Extensive experiments on multiple benchmarks show that CoGenCast achieves competitive performance compared to previous baselines. Code is available at https://github.com/liuyaguo/_CoGenCast.
APA
Cheng, M., Liu, Y., Wang, D., Tao, X. & Liu, Q.. (2026). CoGenCast: A Coupled Autoregressive–Flow Generative Framework for Time Series Forecasting. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:18941-18965 Available from https://proceedings.mlr.press/v306/cheng26h.html.

Related Material