Discrete Flow Matching for Offline-to-Online Reinforcement Learning

• URL: http://arxiv.org/abs/2605.12379v1
• Date: Tue, 12 May 2026 16:44:02 GMT
• Title: Discrete Flow Matching for Offline-to-Online Reinforcement Learning
• Abstract summary: DRIFT is an online fine-tuning method that updates an offline pretrained continuous-time Markov chain (CTMC) policy with an advantage-weighted discrete flow matching loss.<n>For large discrete action spaces, we introduce a candidate-set approximation that updates the actor over a small subset of actions sampled from reference-policy rollouts.<n>Experiments on prevailing discrete action RL task show that our method provides stable offline-to-online improvement across all tasks.
• Score: 10.112779201155005
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Many reinforcement learning (RL) tasks have discrete action spaces, but most generative policy methods based on diffusion and flow matching are designed for continuous control. Meanwhile, generative policies usually rely heavily on offline datasets and offline-to-online RL is itself challenging, as the policy must improve from new interaction without losing useful behavior learned from static data. To address those challenges, we introduce DRIFT, an online fine-tuning method that updates an offline pretrained continuous-time Markov chain (CTMC) policy with an advantage-weighted discrete flow matching loss. To preserve useful pretrained knowledge, we add a path-space penalty that regularizes the full CTMC trajectory distribution, rather than only the final action distribution. For large discrete action spaces, we introduce a candidate-set approximation that updates the actor over a small subset of actions sampled from reference-policy rollouts and uniform exploration. Our theoretical analysis shows that the candidate-set error is controlled by missing target probability mass, and the induced CTMC generator error decreases as the candidate set covers more high-probability actions. Experiments on prevailing discrete action RL task show that our method provides stable offline-to-online improvement across all tasks, achieving the highest average score on Jericho with a simple GRU encoder while outperforming methods that use pretrained language models. Controlled experiments further confirm that the path-space penalty remains bounded during fine-tuning and that the CTMC generator adapts to shifted rewards faster than deterministic baselines. The candidate-set mechanism is supported by a stability analysis showing that the generator error decreases exponentially with candidate coverage.

Related papers

• DISA: Offline Importance Sampling for Distribution-Matching LLM-RL [56.9445657766829]
  We introduce DISA, which moves this calibration problem outside the RL loop.<n> DISA draws proposal trajectories offline, estimates the partition function via importance sampling, and freezes the resulting partition-function estimate.<n>On two open-weight backbones across six math and three code benchmarks, DISA matches or exceeds the online-coupled distribution-matching baseline Flow.
  arXiv  Detail & Related papers  (2026-05-17T07:14:44Z)
• Aligning Flow Map Policies with Optimal Q-Guidance [50.514994916864275]
  Flow map policies are designed for fast action generation by learning to take arbitrary-size jumps.<n>We derive FLOW MAP Q-GUIDANCE (FMQ), a principled closed-form learning target that is optimal for adapting offline flow map policies under a critic-guided trust-region constraint.<n>FMQ achieves state-of-the-art performance in offline-to-online RL, outperforming the previous one-step policy MVP by a relative improvement of 21.3% on the average success rate.
  arXiv  Detail & Related papers  (2026-05-12T17:12:29Z)
• Improving Search Agent with One Line of Code [68.58667107354253]
  Tool-based Agentic Reinforcement Learning (TARL) has emerged as a promising paradigm for training search agents to interact with external tools.<n>We propose textbfSearch textbfAgent textbfPolicy textbfOptimization (textbfSAPO), which stabilizes training via a conditional token-level KL constraint.
  arXiv  Detail & Related papers  (2026-03-10T04:07:39Z)
• Controllable Exploration in Hybrid-Policy RLVR for Multi-Modal Reasoning [88.42566960813438]
  CalibRL is a hybrid-policy RLVR framework that supports controllable exploration with expert guidance.<n>CalibRL increases policy entropy in a guided manner and clarifies the target distribution.<n>Experiments across eight benchmarks, including both in-domain and out-of-domain settings, demonstrate consistent improvements.
  arXiv  Detail & Related papers  (2026-02-22T07:23:36Z)
• TD-JEPA: Latent-predictive Representations for Zero-Shot Reinforcement Learning [63.73629127832652]
  We introduce TD-JEPA, which leverages TD-based latent-predictive representations into unsupervised RL.<n> TD-JEPA trains explicit state and task encoders, a policy-conditioned multi-step predictor, and a set of parameterized policies directly in latent space.<n> Empirically, TD-JEPA matches or outperforms state-of-the-art baselines on locomotion, navigation, and manipulation tasks across 13 datasets.
  arXiv  Detail & Related papers  (2025-10-01T10:21:18Z)
• Flow-Based Single-Step Completion for Efficient and Expressive Policy Learning [0.0]
  We propose a generative policy trained with an augmented flow-matching objective to predict direct completion vectors from intermediate flow samples.<n>Our method scales effectively to offline, offline-to-online, and online RL settings, offering substantial gains in speed and adaptability.<n>We extend SSCP to goal-conditioned RL, enabling flat policies to exploit subgoal structures without explicit hierarchical inference.
  arXiv  Detail & Related papers  (2025-06-26T16:09:53Z)
• Diverse Transformer Decoding for Offline Reinforcement Learning Using Financial Algorithmic Approaches [4.364595470673757]
  Portfolio Beam Search (PBS) is a simple-yet-effective alternative to Beam Search (BS)<n>We develop an uncertainty-aware diversification mechanism, which we integrate into a sequential decoding algorithm at inference time.<n>We empirically demonstrate the effectiveness of PBS on the D4RL benchmark, where it achieves higher returns and significantly reduces outcome variability.
  arXiv  Detail & Related papers  (2025-02-13T15:51:46Z)
• Stabilizing Linear Passive-Aggressive Online Learning with Weighted Reservoir Sampling [46.01254613933967]
  Online learning methods are still highly effective for high-dimensional streaming data, out-of-core processing, and other throughput-sensitive applications. Many such algorithms rely on fast adaptation to individual errors as a key to their convergence. While such algorithms enjoy low theoretical regret, in real-world deployment they can be sensitive to individual outliers that cause the algorithm to over-correct.
  arXiv  Detail & Related papers  (2024-10-31T03:35:48Z)
• CDSA: Conservative Denoising Score-based Algorithm for Offline Reinforcement Learning [25.071018803326254]
  Distribution shift is a major obstacle in offline reinforcement learning. Previous conservative offline RL algorithms struggle to generalize to unseen actions. We propose to use the gradient fields of the dataset density generated from a pre-trained offline RL algorithm to adjust the original actions.
  arXiv  Detail & Related papers  (2024-06-11T17:59:29Z)
• Action-Quantized Offline Reinforcement Learning for Robotic Skill Learning [68.16998247593209]
  offline reinforcement learning (RL) paradigm provides recipe to convert static behavior datasets into policies that can perform better than the policy that collected the data. In this paper, we propose an adaptive scheme for action quantization. We show that several state-of-the-art offline RL methods such as IQL, CQL, and BRAC improve in performance on benchmarks when combined with our proposed discretization scheme.
  arXiv  Detail & Related papers  (2023-10-18T06:07:10Z)
• Offline Policy Optimization in RL with Variance Regularizaton [142.87345258222942]
  We propose variance regularization for offline RL algorithms, using stationary distribution corrections. We show that by using Fenchel duality, we can avoid double sampling issues for computing the gradient of the variance regularizer. The proposed algorithm for offline variance regularization (OVAR) can be used to augment any existing offline policy optimization algorithms.
  arXiv  Detail & Related papers  (2022-12-29T18:25:01Z)
• Regularizing a Model-based Policy Stationary Distribution to Stabilize Offline Reinforcement Learning [62.19209005400561]
  offline reinforcement learning (RL) extends the paradigm of classical RL algorithms to purely learning from static datasets. A key challenge of offline RL is the instability of policy training, caused by the mismatch between the distribution of the offline data and the undiscounted stationary state-action distribution of the learned policy. We regularize the undiscounted stationary distribution of the current policy towards the offline data during the policy optimization process.
  arXiv  Detail & Related papers  (2022-06-14T20:56:16Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.