arXiv cs.AI· Nabuat Zaman Nahim, Fairoz Nower Khan, Peizhong Ju·· 3 天前AI 评分22
强化学习中的流匹配综述
A Survey of Flow Matching in Reinforcement Learning
AI 导读
一篇发表于 TMLR 2026 的 52 页综述系统梳理了强化学习中的流匹配（Flow Matching）方法。该综述沿两个维度组织文献：所建模的目标分布（动作策略、价值评论家、转移动态）与 RL 信号的整合机制（能量加权回归、基于流的策略梯度、组相对策略优化），并对比了高斯与扩散基线，覆盖离散及非欧几里得动作空间等前沿方向。
正文 · 原文
Abstract:Flow Matching (FM) has recently emerged as a principled and efficient generative modeling framework for reinforcement learning (RL), enabling expressive, multimodal policy parameterizations via deterministic probability transport. Compared to diffusion-based policies that rely on stochastic denoising chains, FM uses sampling based on ordinary differential equations (ODEs), with learned velocity fields, which can substantially reduce inference latency and simplify the incorporation of RL objectives. As research in flow-based RL rapidly accelerates across offline continuous control, online fine-tuning, and foundation model alignment, the literature has become highly fragmented. In this survey, we provide a comprehensive taxonomy of flow-matching approaches in reinforcement learning. We organize the literature along two axes: the target distribution being modeled (e.g., action policies, value critics, transition dynamics) and the mechanism of RL signal integration (e.g., energy-weighted regression, flow-based policy gradients, and group relative policy optimization). Furthermore, we survey emerging frontiers such as discrete and non-Euclidean action spaces, provide a systematic comparative analysis against Gaussian and diffusion baselines, and outline critical open problems. Ultimately, this survey serves as a foundational roadmap for the next generation of generative reinforcement learning and alignment.
| Comments: | 52 pages. Published in Transactions on Machine Learning Research (TMLR), 2026. Survey Certification. Reviewed on OpenReview: this https URL |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.16275 [cs.LG] |
| (or arXiv:2610.16275v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.16275 arXiv-issued DOI via DataCite (pending registration) |
|
| Journal reference: | Transactions on Machine Learning Research, 2026 |
Submission history
From: Nabuat Zaman Nahim [view email]
[v1]
Mon, 12 Oct 2026 06:55:17 UTC (93 KB)
来源：arXiv cs.AI · arxiv.org