RAMAC: Multimodal Risk-Aware Offline Reinforcement Learning and the Role of Behavior Regularization

Kai Fukazawa, Kunal Mundada, Iman Soltani
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:32295-32324, 2026.

Abstract

In safety-critical domains where online data collection is infeasible, offline reinforcement learning (RL) is attractive only if policies achieve high returns without catastrophic lower-tail risk. Prior work on risk-averse offline RL achieves safety at the cost of value- or model-based pessimism, and restricted policy classes that limit policy expressiveness, whereas diffusion/flow-based expressive generative policies have largely been used in risk-neutral settings. We introduce Risk-Aware Multimodal Actor-Critic (RAMAC), a simple, modular, model-free framework that couples an expressive generative actor (e.g., diffusion/flow) with a distributional critic and optimizes a composite objective that combines Conditional Value-at-Risk (CVaR) with behavioral cloning (BC), enabling risk-sensitive learning in complex multimodal scenarios. Since out-of-distribution (OOD) actions are a major driver of catastrophic failures in offline RL, we further provide an objective-level analysis showing that controlling behavior divergence via BC suppresses OOD actions and stabilizes CVaR. Instantiating RAMAC with a diffusion actor, we illustrate these insights on a 2-D risky bandit and evaluate on Stochastic-D4RL, observing consistent gains in CVaR$_{0.1}$ while maintaining strong returns.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-fukazawa26a, title = {{RAMAC}: Multimodal Risk-Aware Offline Reinforcement Learning and the Role of Behavior Regularization}, author = {Fukazawa, Kai and Mundada, Kunal and Soltani, Iman}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {32295--32324}, 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/fukazawa26a/fukazawa26a.pdf}, url = {https://proceedings.mlr.press/v306/fukazawa26a.html}, abstract = {In safety-critical domains where online data collection is infeasible, offline reinforcement learning (RL) is attractive only if policies achieve high returns without catastrophic lower-tail risk. Prior work on risk-averse offline RL achieves safety at the cost of value- or model-based pessimism, and restricted policy classes that limit policy expressiveness, whereas diffusion/flow-based expressive generative policies have largely been used in risk-neutral settings. We introduce Risk-Aware Multimodal Actor-Critic (RAMAC), a simple, modular, model-free framework that couples an expressive generative actor (e.g., diffusion/flow) with a distributional critic and optimizes a composite objective that combines Conditional Value-at-Risk (CVaR) with behavioral cloning (BC), enabling risk-sensitive learning in complex multimodal scenarios. Since out-of-distribution (OOD) actions are a major driver of catastrophic failures in offline RL, we further provide an objective-level analysis showing that controlling behavior divergence via BC suppresses OOD actions and stabilizes CVaR. Instantiating RAMAC with a diffusion actor, we illustrate these insights on a 2-D risky bandit and evaluate on Stochastic-D4RL, observing consistent gains in CVaR$_{0.1}$ while maintaining strong returns.} }
Endnote
%0 Conference Paper %T RAMAC: Multimodal Risk-Aware Offline Reinforcement Learning and the Role of Behavior Regularization %A Kai Fukazawa %A Kunal Mundada %A Iman Soltani %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-fukazawa26a %I PMLR %P 32295--32324 %U https://proceedings.mlr.press/v306/fukazawa26a.html %V 306 %X In safety-critical domains where online data collection is infeasible, offline reinforcement learning (RL) is attractive only if policies achieve high returns without catastrophic lower-tail risk. Prior work on risk-averse offline RL achieves safety at the cost of value- or model-based pessimism, and restricted policy classes that limit policy expressiveness, whereas diffusion/flow-based expressive generative policies have largely been used in risk-neutral settings. We introduce Risk-Aware Multimodal Actor-Critic (RAMAC), a simple, modular, model-free framework that couples an expressive generative actor (e.g., diffusion/flow) with a distributional critic and optimizes a composite objective that combines Conditional Value-at-Risk (CVaR) with behavioral cloning (BC), enabling risk-sensitive learning in complex multimodal scenarios. Since out-of-distribution (OOD) actions are a major driver of catastrophic failures in offline RL, we further provide an objective-level analysis showing that controlling behavior divergence via BC suppresses OOD actions and stabilizes CVaR. Instantiating RAMAC with a diffusion actor, we illustrate these insights on a 2-D risky bandit and evaluate on Stochastic-D4RL, observing consistent gains in CVaR$_{0.1}$ while maintaining strong returns.
APA
Fukazawa, K., Mundada, K. & Soltani, I.. (2026). RAMAC: Multimodal Risk-Aware Offline Reinforcement Learning and the Role of Behavior Regularization. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:32295-32324 Available from https://proceedings.mlr.press/v306/fukazawa26a.html.

Related Material