Efficient and Uncertainty-Aware Diffusion Framework for Offline-to-Online Reinforcement Learning

Ha Manh Bui, Metod Jazbec, Eric Nalisnick, Anqi Liu
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:10199-10218, 2026.

Abstract

Offline-to-Online Reinforcement Learning (O2O-RL) leverages an offline, pre-trained policy to minimize costly online interactions. Although data-efficient, O2O-RL is susceptible to shifts between offline and online distributions. Existing work aims to mitigate the harm of this shift by finetuning the policy on trajectory data sampled from a diffusion model. Inspired by this line of work, we propose DUAL: an efficient Diffusion Uncertainty-Aware framework for offline-to-online reinforcement Learning. DUAL utilizes the prior knowledge of the diffusion model to distill a fast-sampling diffusion actor policy and transition model in the offline phase. DUAL also employs a Laplace approximation and distance transition-state-shift detection, thereby using uncertainty quantification to improve exploration versus exploitation in the online phase. We formally show that our actor loss with the Laplace approximation provides a proxy for a principled estimate of epistemic uncertainty. Empirically, DUAL improves the online expected return over O2O-RL baselines across multiple settings and environments.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-bui26a, title = {Efficient and Uncertainty-Aware Diffusion Framework for Offline-to-Online Reinforcement Learning}, author = {Bui, Ha Manh and Jazbec, Metod and Nalisnick, Eric and Liu, Anqi}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {10199--10218}, 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/bui26a/bui26a.pdf}, url = {https://proceedings.mlr.press/v306/bui26a.html}, abstract = {Offline-to-Online Reinforcement Learning (O2O-RL) leverages an offline, pre-trained policy to minimize costly online interactions. Although data-efficient, O2O-RL is susceptible to shifts between offline and online distributions. Existing work aims to mitigate the harm of this shift by finetuning the policy on trajectory data sampled from a diffusion model. Inspired by this line of work, we propose DUAL: an efficient Diffusion Uncertainty-Aware framework for offline-to-online reinforcement Learning. DUAL utilizes the prior knowledge of the diffusion model to distill a fast-sampling diffusion actor policy and transition model in the offline phase. DUAL also employs a Laplace approximation and distance transition-state-shift detection, thereby using uncertainty quantification to improve exploration versus exploitation in the online phase. We formally show that our actor loss with the Laplace approximation provides a proxy for a principled estimate of epistemic uncertainty. Empirically, DUAL improves the online expected return over O2O-RL baselines across multiple settings and environments.} }
Endnote
%0 Conference Paper %T Efficient and Uncertainty-Aware Diffusion Framework for Offline-to-Online Reinforcement Learning %A Ha Manh Bui %A Metod Jazbec %A Eric Nalisnick %A Anqi 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-bui26a %I PMLR %P 10199--10218 %U https://proceedings.mlr.press/v306/bui26a.html %V 306 %X Offline-to-Online Reinforcement Learning (O2O-RL) leverages an offline, pre-trained policy to minimize costly online interactions. Although data-efficient, O2O-RL is susceptible to shifts between offline and online distributions. Existing work aims to mitigate the harm of this shift by finetuning the policy on trajectory data sampled from a diffusion model. Inspired by this line of work, we propose DUAL: an efficient Diffusion Uncertainty-Aware framework for offline-to-online reinforcement Learning. DUAL utilizes the prior knowledge of the diffusion model to distill a fast-sampling diffusion actor policy and transition model in the offline phase. DUAL also employs a Laplace approximation and distance transition-state-shift detection, thereby using uncertainty quantification to improve exploration versus exploitation in the online phase. We formally show that our actor loss with the Laplace approximation provides a proxy for a principled estimate of epistemic uncertainty. Empirically, DUAL improves the online expected return over O2O-RL baselines across multiple settings and environments.
APA
Bui, H.M., Jazbec, M., Nalisnick, E. & Liu, A.. (2026). Efficient and Uncertainty-Aware Diffusion Framework for Offline-to-Online Reinforcement Learning. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:10199-10218 Available from https://proceedings.mlr.press/v306/bui26a.html.

Related Material