Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Doo, JaeHyeok, Jeon, Byeongguk, Ye, Seonghyeon, Lee, Kimin, Seo, Minjoon
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2605.13435
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918499634380800
author Doo, JaeHyeok
Jeon, Byeongguk
Ye, Seonghyeon
Lee, Kimin
Seo, Minjoon
author_facet Doo, JaeHyeok
Jeon, Byeongguk
Ye, Seonghyeon
Lee, Kimin
Seo, Minjoon
contents There is growing interest in utilizing flow-based models as decision-making policies in reinforcement learning due to their high expressive capacity. However, effectively leveraging this expressivity for value maximization remains challenging, as naive gradient-based optimization requires backpropagating through numerical solvers and often leads to instability. Existing approaches typically address this issue by restricting the expressive capacity of flow-based policies, resulting in a trade-off between optimization stability and representational flexibility. To resolve this, we introduce Q-Flow, a framework that leverages the deterministic nature of flow dynamics to explicitly propagate terminal trajectory value to intermediate latent states along the policy-induced flow. This formulation enables stable policy optimization using intermediate value gradients without unrolling the numerical solver, effectively bridging the gap between stability and expressivity. We evaluate Q-Flow in the offline learning setting on the challenging OGBench suite, where it consistently outperforms state-of-the-art baselines by an average of 10.6 percentage points, while also enabling stable online adaptation within the same framework.
format Preprint
id arxiv_https___arxiv_org_abs_2605_13435
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Q-Flow: Stable and Expressive Reinforcement Learning with Flow-Based Policy
Doo, JaeHyeok
Jeon, Byeongguk
Ye, Seonghyeon
Lee, Kimin
Seo, Minjoon
Machine Learning
Artificial Intelligence
There is growing interest in utilizing flow-based models as decision-making policies in reinforcement learning due to their high expressive capacity. However, effectively leveraging this expressivity for value maximization remains challenging, as naive gradient-based optimization requires backpropagating through numerical solvers and often leads to instability. Existing approaches typically address this issue by restricting the expressive capacity of flow-based policies, resulting in a trade-off between optimization stability and representational flexibility. To resolve this, we introduce Q-Flow, a framework that leverages the deterministic nature of flow dynamics to explicitly propagate terminal trajectory value to intermediate latent states along the policy-induced flow. This formulation enables stable policy optimization using intermediate value gradients without unrolling the numerical solver, effectively bridging the gap between stability and expressivity. We evaluate Q-Flow in the offline learning setting on the challenging OGBench suite, where it consistently outperforms state-of-the-art baselines by an average of 10.6 percentage points, while also enabling stable online adaptation within the same framework.
title Q-Flow: Stable and Expressive Reinforcement Learning with Flow-Based Policy
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2605.13435