Offline RL

Generative models have gained significant traction in offline reinforcement learning (RL) due to their ability to model complex trajectory distributions. However, existing generation-based approaches still struggle with long-horizon tasks characterized by sparse rewards. Some hierarchical generation methods have been developed to mitigate this issue by decomposing the original problem into shorter-horizon subproblems using one policy and generating detailed actions with another. While effective, these methods often overlook the multi-scale temporal structure inherent in trajectories, resulting in suboptimal performance. To overcome these limitations, we propose MAGE, a Multi-scale Autoregressive GEneration-based offline RL method. MAGE incorporates a condition-guided multi-scale autoencoder to learn hierarchical trajectory representations, along with a multi-scale transformer that autoregressively generates trajectory representations from coarse to fine temporal scales. MAGE effectively captures temporal dependencies of trajectories at multiple resolutions. Additionally, a condition-guided decoder is employed to exert precise control over short-term behaviors. Extensive experiments on five offline RL benchmarks against fifteen baseline algorithms show that MAGE successfully integrates multi-scale trajectory modeling with conditional guidance, generating coherent and controllable trajectories in long-horizon sparse-reward settings.

Offline goal-conditioned reinforcement learning (GCRL) learns goal-reaching behaviors from static datasets, but accurate value estimation remains challenging under limited state-action coverage. Existing physics-informed approaches address this by imposing pointwise distance-like geometric constraints derived from Hamilton--Jacobi--Bellman (HJB) optimality principles, often through first-order partial differential equations such as the Eikonal equation. However, enforcing local consistency through explicit differential structure can become unstable in complex, high-dimensional environments. Our key insight is to instead reinterpret distance-like constraints as an expectation over a local spatial measure. By aggregating constraints over this measure rather than evaluating them pointwise, the objective acts as a spatial mollifier, inducing distance-like value geometry without requiring expensive differential operators. We refer to this as Mollified Value Learning (MVL). Experiments across navigation and high-dimensional robotic manipulation tasks show that MVL learns structured, value representations, improving goal-reaching performance, when used with implicit value representation learning methods. Open-source codes are available at https://github.com/HrishikeshVish/MVL.

Off-dynamics offline reinforcement learning (RL) aims to learn a policy for a target domain using limited target data and abundant source data collected under different transition dynamics. Existing methods typically address dynamics mismatch either globally over the state space or via pointwise data filtering; these approaches can miss localized cross-domain similarities or incur high computational cost. We propose Localized Dynamics-Aware Domain Adaptation (LoDADA), which exploits localized dynamics mismatch to better reuse source data. LoDADA clusters transitions from source and target datasets and estimates cluster-level dynamics discrepancy via domain discrimination. Source transitions from clusters with small discrepancy are retained, while those from clusters with large discrepancy are filtered out. This yields a fine-grained and scalable data selection strategy that avoids overly coarse global assumptions and expensive per-sample filtering. We provide theoretical insights and extensive experiments across environments with diverse global and local dynamics shifts. Results show that LoDADA consistently outperforms state-of-the-art off-dynamics offline RL methods by better leveraging localized distribution mismatch.

Diffusion models have shown promising capabilities in trajectory generation for planning in offline reinforcement learning (RL). However, conventional diffusion-based planning methods often fail to account for the fact that generating trajectories in RL requires unique consistency between transitions to ensure coherence in real environments. This oversight can result in considerable discrepancies between the generated trajectories and the underlying mechanisms of a real environment. To address this problem, we propose a novel diffusion-based planning method, termed as Diffusion Modulation via Environment Mechanism Modeling (DMEMM). DMEMM modulates diffusion model training by incorporating key RL environment mechanisms, particularly transition dynamics and reward functions. Experimental results demonstrate that DMEMM achieves state-of-the-art performance for planning with offline reinforcement learning.

Offline reinforcement learning (RL) has garnered significant interest due to its safe and easily scalable paradigm. However, training under this paradigm presents its own challenge: the extrapolation error stemming from out-of-distribution (OOD) data. Existing methodologies have endeavored to address this issue through means like penalizing OOD Q-values or imposing similarity constraints on the learned policy and the behavior policy. Nonetheless, these approaches are often beset by limitations such as being overly conservative in utilizing OOD data, imprecise OOD data characterization, and significant computational overhead. To address these challenges, this paper introduces an Uncertainty-Aware Rank-One Multi-Input Multi-Output (MIMO) Q Network framework. The framework aims to enhance Offline Reinforcement Learning by fully leveraging the potential of OOD data while still ensuring efficiency in the learning process. Specifically, the framework quantifies data uncertainty and harnesses it in the training losses, aiming to train a policy that maximizes the lower confidence bound of the corresponding Q-function. Furthermore, a Rank-One MIMO architecture is introduced to model the uncertainty-aware Q-function, \TP{offering the same ability for uncertainty quantification as an ensemble of networks but with a cost nearly equivalent to that of a single network}. Consequently, this framework strikes a harmonious balance between precision, speed, and memory efficiency, culminating in improved overall performance. Extensive experimentation on the D4RL benchmark demonstrates that the framework attains state-of-the-art performance while remaining computationally efficient. By incorporating the concept of uncertainty quantification, our framework offers a promising avenue to alleviate extrapolation errors and enhance the efficiency of offline RL.

Mamba-based models have drawn much attention in offline RL. However, their selective mechanism often detrimental when key steps in RL sequences are omitted. To address these issues, we propose a simple yet effective structure, called Decision MetaMamba (DMM), which replaces Mamba's token mixer with a dense layer-based sequence mixer and modifies positional structure to preserve local information. By performing sequence mixing that considers all channels simultaneously before Mamba, DMM prevents information loss due to selective scanning and residual gating. Extensive experiments demonstrate that our DMM delivers the state-of-the-art performance across diverse RL tasks. Furthermore, DMM achieves these results with a compact parameter footprint, demonstrating strong potential for real-world applications.

Planning-based reinforcement learning for continuous control is bottlenecked by two practical issues: planning at primitive time scales leads to prohibitive branching and long horizons, while real environments are frequently partially observable and exhibit regime shifts that invalidate stationary, fully observed dynamics assumptions. We introduce I-TAP (In-Context Latent Temporal-Abstraction Planner), an offline RL framework that unifies in-context adaptation with online planning in a learned discrete temporal-abstraction space. From offline trajectories, I-TAP learns an observation-conditioned residual-quantization VAE that compresses each observation-macro-action segment into a coarse-to-fine stack of discrete residual tokens, and a temporal Transformer that autoregressively predicts these token stacks from a short recent history. The resulting sequence model acts simultaneously as a context-conditioned prior over abstract actions and a latent dynamics model. At test time, I-TAP performs Monte Carlo Tree Search directly in token space, using short histories for implicit adaptation without gradient update, and decodes selected token stacks into executable actions. Across deterministic MuJoCo, stochastic MuJoCo with per-episode latent dynamics regimes, and high-dimensional Adroit manipulation, including partially observable variants, I-TAP consistently matches or outperforms strong model-free and model-based offline baselines, demonstrating efficient and robust in-context planning under stochastic dynamics and partial observability.

The dataset distributions in offline reinforcement learning (RL) often exhibit complex and multi-modal distributions, necessitating expressive policies to capture such distributions beyond widely-used Gaussian policies. To handle such complex and multi-modal datasets, in this paper, we propose Flow Actor-Critic, a new actor-critic method for offline RL, based on recent flow policies. The proposed method not only uses the flow model for actor as in previous flow policies but also exploits the expressive flow model for conservative critic acquisition to prevent Q-value explosion in out-of-data regions. To this end, we propose a new form of critic regularizer based on the flow behavior proxy model obtained as a byproduct of flow-based actor design. Leveraging the flow model in this joint way, we achieve new state-of-the-art performance for test datasets of offline RL including the D4RL and recent OGBench benchmarks.

The paradigm of decision-making has been revolutionised by reinforcement learning and deep learning. Although this has led to significant progress in domains such as robotics, healthcare, and finance, the use of RL in practice is challenging, particularly when learning decision policies in high-stakes applications that may require guarantees. Traditional RL algorithms rely on a large number of online interactions with the environment, which is problematic in scenarios where online interactions are costly, dangerous, or infeasible. However, learning from offline datasets is hindered by the presence of hidden confounders. Such confounders can cause spurious correlations in the dataset and can mislead the agent into taking suboptimal or adversarial actions. Firstly, we address the problem of learning from offline datasets in the presence of hidden confounders. We work with instrumental variables (IVs) to identify the causal effect, which is an instance of a conditional moment restrictions (CMR) problem. Inspired by double/debiased machine learning, we derive a sample-efficient algorithm for solving CMR problems with convergence and optimality guarantees, which outperforms state-of-the-art algorithms. Secondly, we relax the conditions on the hidden confounders in the setting of (offline) imitation learning, and adapt our CMR estimator to derive an algorithm that can learn effective imitator policies with convergence rate guarantees. Finally, we consider the problem of learning high-level objectives expressed in linear temporal logic (LTL) and develop a provably optimal learning algorithm that improves sample efficiency over existing methods. Through evaluation on reinforcement learning benchmarks and synthetic and semi-synthetic datasets, we demonstrate the usefulness of the methods developed in this thesis in real-world decision making.

Modern offline Reinforcement Learning (RL) methods find performant actor-critics, however, fine-tuning these actor-critics online with value-based RL algorithms typically causes immediate drops in performance. We provide evidence consistent with the hypothesis that, in the loss landscape, offline maxima for prior algorithms and online maxima are separated by low-performance valleys that gradient-based fine-tuning traverses. Following this, we present Score Matched Actor-Critic (SMAC), an offline RL method designed to learn actor-critics that transition to online value-based RL algorithms with no drop in performance. SMAC avoids valleys between offline and online maxima by regularizing the Q-function during the offline phase to respect a first-order derivative equality between the score of the policy and action-gradient of the Q-function. We experimentally demonstrate that SMAC converges to offline maxima that are connected to better online maxima via paths with monotonically increasing reward found by first-order optimization. SMAC achieves smooth transfer to Soft Actor-Critic and TD3 in 6/6 D4RL tasks. In 4/6 environments, it reduces regret by 34-58% over the best baseline.

Offline safe reinforcement learning (RL) is increasingly important for cyber-physical systems (CPS), where safety violations during training are unacceptable and only pre-collected data are available. Existing offline safe RL methods typically balance reward-safety tradeoffs through constraint relaxation or joint optimization, but they often lack structural mechanisms to prevent safety drift. We propose LexiSafe, a lexicographic offline RL framework designed to preserve safety-aligned behavior. We first develop LexiSafe-SC, a single-cost formulation for standard offline safe RL, and derive safety-violation and performance-suboptimality bounds that together yield sample-complexity guarantees. We then extend the framework to hierarchical safety requirements with LexiSafe-MC, which supports multiple safety costs and admits its own sample-complexity analysis. Empirically, LexiSafe demonstrates reduced safety violations and improved task performance compared to constrained offline baselines. By unifying lexicographic prioritization with structural bias, LexiSafe offers a practical and theoretically grounded approach for safety-critical CPS decision-making.

Preference-based reinforcement learning (PbRL) can help avoid sophisticated reward designs and align better with human intentions, showing great promise in various real-world applications. However, obtaining human feedback for preferences can be expensive and time-consuming, which forms a strong barrier for PbRL. In this work, we address the problem of low query efficiency in offline PbRL, pinpointing two primary reasons: inefficient exploration and overoptimization of learned reward functions. In response to these challenges, we propose a novel algorithm, Offline PbRL via In-Dataset Exploration (OPRIDE), designed to enhance the query efficiency of offline PbRL. OPRIDE consists of two key features: a principled exploration strategy that maximizes the informativeness of the queries and a discount scheduling mechanism aimed at mitigating overoptimization of the learned reward functions. Through empirical evaluations, we demonstrate that OPRIDE significantly outperforms prior methods, achieving strong performance with notably fewer queries. Moreover, we provide theoretical guarantees of the algorithm's efficiency. Experimental results across various locomotion, manipulation, and navigation tasks underscore the efficacy and versatility of our approach.

Mobile edge computing (MEC) based wireless metaverse services offer an untethered, immersive experience to users, where the superior quality of experience (QoE) needs to be achieved under stringent latency constraints and visual quality demands. To achieve this, MEC-based intelligent resource allocation for virtual reality users needs to be supported by coordination across MEC servers to harness distributed data. Federated learning (FL) is a promising solution, and can be combined with reinforcement learning (RL) to develop generalized policies across MEC-servers. However, conventional FL incurs transmitting the full model parameters across the MEC-servers and the cloud, and suffer performance degradation due to naive global aggregation, especially in heterogeneous multi-radio access technology environments. To address these challenges, this paper proposes Federated Split Decision Transformer (FSDT), an offline RL framework where the transformer model is partitioned between MEC servers and the cloud. Agent-specific components (e.g., MEC-based embedding and prediction layers) enable local adaptability, while shared global layers in the cloud facilitate cooperative training across MEC servers. Experimental results demonstrate that FSDT enhances QoE for up to 10% in heterogeneous environments compared to baselines, while offloadingnearly 98% of the transformer model parameters to the cloud, thereby reducing the computational burden on MEC servers.

Offline reinforcement learning (RL) offers a promising framework for deploying autonomous systems in safety-critical settings without the risks of online exploration. However, learning policies that simultaneously achieve high performance and strong safety guarantees from fixed datasets remains a fundamental challenge. Many existing safe offline RL approaches typically rely on soft constraint formulations, which may permit safety violations and are sensitive to distributional shift. In contrast, formal methods such as Hamilton--Jacobi (HJ) reachability and Control Barrier Functions (CBFs) provide rigorous safety guarantees, but often yield overly conservative solutions, often neglecting performance. In this work, we bridge this gap by formulating safe offline RL as a state-constrained optimal control problem, where safety is enforced through hard state constraints and performance is captured via a reward function. The resulting value function satisfies a Hamilton--Jacobi--Bellman (HJB) equation, which we approximate using offline RL on fixed datasets by learning an epigraph feasibility value function and extracting policies via flow matching. This formulation enables principled integration of safety guarantees with data-driven policy optimization. Empirically, across safety-critical benchmarks including boat navigation, Safety-Gymnasium, and Bullet-Safety-Gym tasks, our approach achieves competitive returns while exhibiting near-zero constraint violations, demonstrating a favorable balance between safety and performance in the offline setting.

Reinforcement learning (RL) is a type of artificial intelligence for making optimal choices. In healthcare, researchers generally use offline RL (ORL), where models are trained and evaluated from retrospective observational data. To accommodate naturally irregular clinical records, researchers often resample the data into uniform time intervals before training (known as binning). However, binning may create counterfactual trajectories that do not represent clinical practice. We first quantified these distortions in patients receiving continuous insulin infusions in the MIMIC-IV intensive care database. Binning reversed the apparent causal order for over 70% of paired glucose measurements and insulin adjustments within the same interval, and frequently masked significant dysglycaemic events and merged distinct treatment decisions together. We then conducted an in silico clinical trial using 30 virtual patients with type 1 diabetes from the FDA-approved UVA/Padova simulator. The simulator was modified to include stochastic intervals between decisions and used to generate a training dataset for offline RL. We trained three ORL algorithms on both the unprocessed dataset and equivalent datasets resampled at 10-minute, 2-hour, and 4-hour intervals. When deployed back into the simulated environment, temporal resampling was found to reduce model performance by up to 60% relative to unprocessed data, with 4-hour binning causing all agents to perform worse than the dataset's baseline. Retrospective evaluation on resampled data actively obscured this effect, predicting 1.5-3x better returns than agents achieved in practice. We recommend that future research in this area prioritises datasets with natural clinical timings between decisions, which may be a necessary step before these models can be safely deployed into patient care.

Generative policies based on diffusion models and flow matching have shown strong promise for offline reinforcement learning (RL), but their applicability remains largely confined to continuous action spaces. To address a broader range of offline RL settings, we extend flow matching to a general framework that supports discrete action spaces with multiple objectives. Specifically, we replace continuous flows with continuous-time Markov chains, trained using a Q-weighted flow matching objective. We then extend our design to multi-agent settings, mitigating the exponential growth of joint action spaces via a factorized conditional path. We theoretically show that, under idealized conditions, optimizing this objective recovers the optimal policy. Extensive experiments further demonstrate that our method performs robustly across diverse settings and benchmarks, including high-dimensional control, multi-agent games, and dynamically changing preferences over multiple objectives, while outperforming traditional offline RL methods in practical multi-modal decision-making scenarios. Our discrete framework can also be applied to continuous-control problems through action quantization, providing a flexible trade-off between representational complexity and performance.

Offline goal-conditioned reinforcement learning (GCRL) is typically benchmarked by the best tuned success rate of each method. This score measures attainable performance, but it does not reveal how reliably a learned goal-conditioned signal can be extracted into a policy: a method could succeed across many value-learning and extraction settings, or only at a narrow, hard-to-find configuration. We study this gap across four methods, GCIQL, GCIVL, QRL, and CRL, under a shared advantage-weighted regression (AWR) extractor. For each method, we construct trainability landscapes over the optimizer learning rate, which affects value learning and actor optimization, and AWR temperature, which controls how selectively the actor imitates high-advantage transitions. Across AntMaze, Cube, and Scene, we observe distinct regimes: high-scoring methods may be broadly accessible or brittle, while broad relative basins may still sit below low absolute ceilings. To interpret these differences, we pair landscapes with post-hoc diagnostics of future-vs-random goal discrimination and AWR weight concentration. Their relationship to downstream success is task-dependent. On AntMaze, where future goals align with path-like progress, these diagnostics explain landscape regimes. On Cube and Scene, goal ranking and manipulation control decouple: methods can rank goals well while failing downstream, or succeed through action-conditioned advantages despite weak future-vs-random separation. These results show that peak tuned success alone does not establish broadly extractable goal-conditioned behavior. Trainability landscapes expose this gap, while extraction diagnostics offer a lower-cost lens on how learned signals become policies.

Training large language models (LLMs) for multi-turn code generation typically relies on online reinforcement learning (RL), in which models learn from repeatedly generated online rollouts. Although effective, online RL can be costly and unstable. In this paper, we present contextual bandit learning with offline trajectories (COBALT), a simple method that reuses offline multi-turn trajectories to efficiently train LLMs for multi-turn code generation. Building on the observation that multi-turn code generation is a one-step recoverable Markov decision process, COBALT first collects code generation trajectories using a reference LLM and divides them into partial trajectories. Then, during online bandit learning, the LLM is trained to complete each partial trajectory through single-step code generation. On LiveCodeBench, COBALT substantially improves R1-Distill 8B and Qwen3 8B by up to 9.0 and 6.2 absolute Pass@1 points and matches the performance of two multi-turn online RL baselines using GRPO and VeRPO. We also perform a perturbation-based analysis of in-context reward hacking and show that Cobalt can improve robustness to misleading execution feedback. Overall, our results demonstrate that Cobalt is a promising solution for multi-turn code generation. Our code and data are available at https://github.com/OSU-NLP-Group/cobalt.

Offline goal-conditioned reinforcement learning remains challenging for long-horizon tasks. While hierarchical approaches mitigate this issue by decomposing tasks, most existing methods rely on separate high- and low-level networks and generate only a single intermediate subgoal, leaving several structural limitations in long-horizon decision-making. To address this limitation, we draw inspiration from chain-of-thought reasoning and propose the Chain-of-Goals Hierarchical Policy (CoGHP), a novel framework that reformulates hierarchical decision-making as autoregressive sequence modeling within a unified architecture. Given a state and a final goal, CoGHP autoregressively generates a sequence of latent subgoals followed by the primitive action, where each latent subgoal acts as a reasoning step that conditions subsequent predictions. To implement this efficiently, we introduce an MLP-Mixer backbone, which supports cross-token communication and captures structural relationships among state, goal, latent subgoals, and action. Across challenging navigation and manipulation benchmarks, CoGHP consistently outperforms strong offline baselines, demonstrating improved performance on long-horizon tasks. Project page: https://wlsdn9350.github.io/projects/coghp/

Learning a common policy from multiple source environments requires accounting for cross-source heterogeneity, and the offline setting introduces an additional challenge of limited data coverage. We formulate this problem as a distributionally robust Markov decision process under a shared linear MDP representation with source-specific transition and reward components. The uncertainty set is constructed through feature-wise mixtures of these components, yielding a $d$-rectangular structure that captures cross-source heterogeneity and admits a robust Bellman recursion. Building on this formulation, we develop Group-Robust Pessimistic Value Iteration (GR-PEVI), an offline policy-learning algorithm that combines feature-wise worst-case aggregation with a data-dependent pessimism penalty. Under a robust partial coverage condition, we establish finite-sample suboptimality guarantees for GR-PEVI and extend the analysis to clustered sources, characterizing the trade-off between the statistical gain from pooling similar sources and the approximation error induced by within-cluster heterogeneity. Simulation studies demonstrate the effectiveness of GR-PEVI, with empirical behavior consistent with the theoretical analysis. We apply GR-PEVI to a multi-source electronic health record study of multiple sclerosis treatment decisions to illustrate its use in a clinical setting.

We study offline reinforcement learning of style-conditioned policies using explicit style supervision via subtrajectory labeling functions. In this setting, aligning style with high task performance is particularly challenging due to distribution shift and inherent conflicts between style and reward. Existing methods, despite introducing numerous definitions of style, often fail to reconcile these objectives effectively. To address these challenges, we propose a unified definition of behavior style and instantiate it into a practical framework. Building on this, we introduce Style-Conditioned Implicit Q-Learning (SCIQL), which leverages offline goal-conditioned RL techniques, such as hindsight relabeling and value learning, and combine it with a new Gated Advantage Weighted Regression mechanism to efficiently optimize task performance while preserving style alignment. Experiments demonstrate that SCIQL achieves superior performance on both objectives compared to prior offline methods. Code, datasets and visuals are available in: https://mathieu-petitbois.github.io/projects/sciql/.

We present DeFlow, a decoupled offline RL framework that leverages flow matching to faithfully capture complex behavior manifolds. Optimizing generative policies is computationally prohibitive, typically necessitating backpropagation through ODE solvers. We address this by learning a lightweight refinement module within an explicit, data-derived trust region of the flow manifold, rather than sacrificing the iterative generation capability via single-step distillation. This way, we bypass solver differentiation and eliminate the need for balancing loss terms, ensuring stable improvement while fully preserving the flow's iterative expressivity. Empirically, DeFlow achieves superior performance on the challenging OGBench benchmark and demonstrates efficient offline-to-online adaptation.

Fitted $Q$-iteration (FQI) is a standard regression-based method for optimal control in offline reinforcement learning, but its stability under function approximation often relies on Bellman completeness, which requires Bellman images of the fitted class to remain in the class. We study Kullback--Leibler (KL)-regularized, or soft, FQI relative to a fixed reference policy without this assumption. Our key insight is that soft control locally inherits the contraction of policy evaluation in a discounted-occupancy norm. At the soft-optimal fixed point, the linearization of the soft Bellman operator is exactly the Bellman operator for the soft-optimal policy, which contracts in its discounted-occupancy norm; projection in the same norm preserves this contraction. Standard soft FQI instead projects under the offline state-action distribution and need not preserve this property. Motivated by this observation, we propose occupancy-reweighted soft FQI, which retains standard Bellman targets and least-squares updates while reweighting regressions by discounted-occupancy ratios induced by the current soft policy. Under $Q$-function realizability and local regularity, we establish local contraction and finite-sample convergence with estimated ratios, without Bellman completeness. We then use temperature annealing to convert the local result into global convergence from arbitrary initialization: sufficiently high temperature provides a globally contractive starting regime, while gradual cooling connects successive local contraction regions to any prescribed positive target temperature. Under an action-gap margin condition, switching at a fixed positive temperature to hard FQI with refreshed occupancy weights also yields population and finite-sample convergence to the unregularized optimum.

Continuous-time stochastic processes underlie many natural and engineered systems. In healthcare, autonomous driving, and industrial control, direct interaction with the environment is often unsafe or impractical, motivating offline reinforcement learning from historical data. However, there is limited statistical understanding of the approximation errors inherent in learning policies from offline datasets. We address this by linking reinforcement learning to the Hamilton-Jacobi-Bellman equation and proposing an operator-theoretic algorithm based on a simple dynamic programming recursion. Specifically, we represent our world model in terms of the infinitesimal generator of controlled diffusion processes learned in a reproducing kernel Hilbert space. By integrating statistical learning methods and operator theory, we establish global convergence of the value function under sufficiently small generator-approximation error and derive explicit finite-sample guarantees with bounds tied directly to system properties such as smoothness and stability. Our theoretical and proof-of-concept results indicate that operator-based approaches may hold promise for solving offline reinforcement learning using continuous-time optimal control.

Offline-to-online reinforcement learning (RL) has emerged as a practical paradigm that leverages offline datasets for pretraining and online interactions for fine-tuning. However, its empirical behavior is highly inconsistent: design choices of online fine-tuning that work well in one setting can fail completely in another. Guided by the stability--plasticity principle, we propose a framework that can explain this inconsistency: We argue that efficient fine-tuning must preserve the utility of the stronger offline prior, whether that is the pretrained policy or the offline dataset, while maintaining sufficient plasticity. This perspective identifies three regimes of online fine-tuning, each requiring distinct stability properties. We validate this framework through a large-scale empirical study, finding that the results strongly align with its predictions in 45 out of 63 cases, with only 3 opposite mismatches. This work provides a framework for guiding design choices in offline-to-online RL based on the relative performance of the offline dataset and the pretrained policy.

Target-conditioned sequence models provide a simple interface for controllable offline decision making, but the requested target return can be an unreliable control signal, especially when the target return lies in underrepresented regions of the dataset. This paper proposes Doctor, a hybrid sequence modeling and reinforced verification framework for controllable target-conditioned offline decision making. Doctor trains a shared masked trajectory Transformer with two complementary objectives: masked trajectory reconstruction for candidate generation and in-sample value learning for action-value verification. At inference time, the model samples multiple nearby target returns, generates candidate actions in parallel, and selects the action whose verified value is closest to the requested target return. We analyze this verifier-guided selection rule and show that its value-level alignment error is bounded by candidate-value coverage around the target return and verifier accuracy. Experiments on D4RL and EpiCare show that Doctor improves target-return alignment under reduced high-return coverage, remains competitive on standard offline return-maximization benchmarks, and enables a single policy to modulate between conservative and aggressive operating points in a simulated clinical decision-making task. These results suggest that reinforced verification can improve the controllability of target-conditioned policies.

Offline reinforcement learning (RL) learns policies from fixed datasets, thereby avoiding costly or unsafe environment interactions. However, its reliance on finite static datasets inherently restricts the ability to generalize beyond the training distribution. Prior solutions based on synthetic data augmentation often fail to generalize to unseen scenarios in the (augmented) dataset. To address these challenges, we propose Retrieval High-quAlity Demonstrations (RAD) for decision-making, which innovatively introduces a retrieval mechanism into offline RL. Specifically, RAD retrieves high-return and reachable states from the offline dataset as target states, and leverages a generative model to generate sub-trajectories conditioned on these targets for planning. Since the targets are high-return states, once the agent reaches such a target, it can continue to obtain high returns by following the associated high-return actions, thereby improving policy generalization. Extensive experiments confirm that RAD achieves competitive or superior performance compared to baselines across diverse benchmarks, validating its effectiveness. Our code is available at https://anonymous.4open.science/r/RAD_0925_1-690E.

Proficient game agents with diverse play styles enrich the gaming experience and enhance the replay value of games. However, recent advancements in game AI based on reinforcement learning have predominantly focused on improving proficiency, whereas methods based on evolution algorithms generate agents with diverse play styles but exhibit subpar performance compared to RL methods. To address this gap, this paper proposes Mixed Proximal Policy Optimization (MPPO), a method designed to improve the proficiency of existing suboptimal agents while retaining their distinct styles. MPPO unifies loss objectives for both online and offline samples and introduces an implicit constraint to approximate demonstrator policies by adjusting the empirical distribution of samples. Empirical results across environments of varying scales demonstrate that MPPO achieves proficiency levels comparable to, or even superior to, pure online algorithms while preserving demonstrators' play styles. This work presents an effective approach for generating highly proficient and diverse game agents, ultimately contributing to more engaging gameplay experiences.

Preference-based Reinforcement Learning (PbRL) entails a variety of approaches for aligning models with human intent to alleviate the burden of reward engineering. However, most previous PbRL work has not investigated the robustness to labeler errors, inevitable with labelers who are non-experts or operate under time constraints. We introduce Similarity as Reward Alignment (SARA), a simple contrastive framework that is both resilient to noisy labels and adaptable to diverse feedback formats. SARA learns a latent representation of preferred samples and computes rewards as similarities to the learned latent. On preference data with varying realistic noise rates, we demonstrate competitive and more stable performance on continuous control offline RL benchmarks, with statistically significant improvements over baselines (Wilcoxon signed-rank, p < 0.01). We also compute correlation to the environment rewards as a proxy for measuring alignment to the underlying preference criteria. We show that the SARA computed rewards display higher correlation across noise rates compared to baselines.

Offline RL (ORL) promises safe and sample-efficient deployment but existing methods rely on undocumented online interactions for hyperparameter tuning and lack reliable fully offline estimates of initial online performance. We introduce SOReL, a fully offline Bayesian model-based RL method that learns a posterior over dynamics, estimates policy value via predictive uncertainty, and enables complete offline hyperparameter selection. We further propose TOReL, which extends this tuning framework to arbitrary model-free and model-based ORL algorithms. We provide a regret analysis showing that Bayesian offline RL achieves the minimax-optimal parametric rate under standard regularity conditions. Together, our methods establish a practical and theoretically grounded framework for fully offline RL.