Learning meaningful abstract models of Markov Decision Processes (MDPs) is crucial for improving generalization from limited data. In this work, we show how geometric priors can be imposed on the low-dimensional representation manifold of a learned transition model. We incorporate known symmetric structures via appropriate choices of the latent space and the associated group actions, which encode prior knowledge about invariances in the environment. In addition, our framework allows the embedding of additional unstructured information alongside these symmetries. We show experimentally that this leads to better predictions of the latent transition model than fully unstructured approaches, as well as better learning on downstream RL tasks, in environments with rotational and translational features, including in first-person views of 3D environments. Additionally, our experiments show that this leads to simpler and more disentangled representations. The full code is available on GitHub to ensure reproducibility.

Reinforcement learning (RL) has driven breakthroughs in AI, from game-play to scientific discovery and AI alignment. However, its broader applicability remains limited by challenges such as low data efficiency and poor generalizability. Recent advances suggest that large language models, with their rich world knowledge and reasoning capabilities, could complement RL by enabling semantic state modeling and task-agnostic planning. In this work, we propose the Agentic Episodic Control (AEC), a novel architecture that integrates RL with LLMs to enhance decision-making. The AEC can leverage a large language model (LLM) to map the observations into language-grounded embeddings, which further can be stored in an episodic memory for rapid retrieval of high-value experiences. Simultaneously, a World-Graph working memory module is utilized to capture structured environmental dynamics in order to enhance relational reasoning. Furthermore, a lightweight critical state detector dynamically arbitrates between the episodic memory recall and the world-model-guided exploration. On the whole, by combining the trial-and-error learning scheme with LLM-derived semantic priors, the proposed AEC can improve both data efficiency and generalizability in reinforcement learning. In experiments on BabyAI-Text benchmark tasks, AEC demonstrates substantial improvements over existing baselines, especially on complex and generalization tasks like FindObj, where it outperforms the best baseline by up to 76%. The proposed AEC framework bridges the strengths of numeric reinforcement learning and symbolic reasoning, which provides a pathway toward more adaptable and sample-efficient agents.

Abstract--The Low Altitude Economy (LAE) network, with its transformative capabilities, is a candidate to become one of the major technological developments of the next decade for air mobility. However, the expected unprecedented density, mobility, and heterogeneity pose challenges and require new approaches, as it renders traditional rule-based approaches inadequate. T o address these challenges, this study introduces artificial intelligence (AI)-based approaches and validation frameworks for transitioning AI-enabled technologies from simulation-based studies to practical and deployable systems. First, AI-based spectrum sensing and coexistence utilizing the distributed nature of LAE nodes is introduced. Then, joint resource allocation and trajectory optimization driven by reinforcement learning is discussed. Bridging the gap between simulation and deployment through experimental platforms such as Aerial Experiments and Research Platform for Advanced Wireless (AERPA W), which are critical for validating models under realistic and non-stationary airspace conditions, is also addressed. The study concludes by highlighting open issues and outlining a forward-looking roadmap for the development of efficient, interoperable, and scalable AI-driven LAE ecosystems. The Low Altitude Economy (LAE) network is poised to become one of the defining technological trends of the next decade. Encompassing the use of the airspace below 3000 metres for economic, social, and operational activities, LAE covers various applications: urban air mobility (e.g., air taxis, emergency medical deliveries), precision agriculture, environmental sensing, surveillance, and logistics, as illustrated ixn Figure 1. M. Chen is with the Department of Electrical and Computer Engineering and Frost Institute for Data Science and Computing, University of Miami, Coral Gables, FL, 33146, USA (email: mingzhe.chen@miami.edu). This work is supported by the NSERC award ALLRP 579869-22 in Canada and the NSF awards CNS-2332834 and CNS-2332835 in the United States.

Diffusion models have emerged as powerful generative frameworks, creating data samples by progressively denoising an initial random state. Traditionally, this initial state is sampled from a simple, fixed distribution like isotropic Gaussian, inherently lacking structure and a direct mechanism for external control. While recent efforts have explored ways to introduce controllability into the diffusion process, particularly at the initialization stage, they often rely on deterministic or heuristic approaches. These methods can be suboptimal, lack expressiveness, and are difficult to scale or integrate into more sophisticated optimization frameworks. In this paper, we introduce NoiseAR, a novel method for AutoRegressive Initial Noise Prior for Diffusion Models. Instead of a static, unstructured source, NoiseAR learns to generate a dynamic and controllable prior distribution for the initial noise. We formulate the generation of the initial noise prior's parameters as an autoregressive probabilistic modeling task over spatial patches or tokens. This approach enables NoiseAR to capture complex spatial dependencies and introduce learned structure into the initial state. Crucially, NoiseAR is designed to be conditional, allowing text prompts to directly influence the learned prior, thereby achieving fine-grained control over the diffusion initialization. Our experiments demonstrate that NoiseAR can generate initial noise priors that lead to improved sample quality and enhanced consistency with conditional inputs, offering a powerful, learned alternative to traditional random initialization. A key advantage of NoiseAR is its probabilistic formulation, which naturally supports seamless integration into probabilistic frameworks like Markov Decision Processes and Reinforcement Learning. Our code will be available at https://github.com/HKUST-SAIL/NoiseAR/

In the context of Federated Reinforcement Learning (FRL), applying Proximal Policy Optimization (PPO) faces challenges related to the update order of its actor and critic due to the aggregation step occurring between successive iterations. In particular, when local actors are updated based on local critic estimations, the algorithm becomes vulnerable to data heterogeneity. As a result, the conventional update order in PPO (critic first, then actor) may cause heterogeneous gradient directions among clients, hindering convergence to a globally optimal policy. To address this issue, we propose FedRAC, which reverses the update order (actor first, then critic) to eliminate the divergence of critics from different clients. Theoretical analysis shows that the convergence bound of FedRAC is immune to data heterogeneity under mild conditions, i.e., bounded level of heterogeneity and accurate policy evaluation. Empirical results indicate that the proposed algorithm obtains higher cumulative rewards and converges more rapidly in five experiments, including three classical RL environments and a highly heterogeneous autonomous driving scenario using the SUMO traffic simulator.

Slow Feature Analysis is a unsupervised representation learning method that extracts slowly varying features from temporal data and can be used as a basis for subsequent reinforcement learning. Often, the behavior that generates the data on which the representation is learned is assumed to be a uniform random walk. Less research has focused on using samples generated by goal-directed behavior, as commonly the case in a reinforcement learning setting, to learn a representation. In a spatial setting, goal-directed behavior typically leads to significant differences in state occupancy between states that are close to a reward location and far from a reward location. Through the perspective of optimal slow features on ergodic Markov chains, this work investigates the effects of these differences on value-function approximation in an idealized setting. Furthermore, three correction routes, which can potentially alleviate detrimental scaling effects, are evaluated and discussed. In addition, the special case of goal-averse behavior is considered.

Recent advances in diffusion-based generative modeling have demonstrated significant promise in tackling long-horizon, sparse-reward tasks by leveraging offline datasets. While these approaches have achieved promising results, their reliability remains inconsistent due to the inherent stochastic risk of producing infeasible trajectories, limiting their applicability in safety-critical applications. We identify that the primary cause of these failures is inaccurate guidance during the sampling procedure, and demonstrate the existence of manifold deviation by deriving a lower bound on the guidance gap. To address this challenge, we propose Local Manifold Approximation and Projection (LoMAP), a training-free method that projects the guided sample onto a low-rank subspace approximated from offline datasets, preventing infeasible trajectory generation. We validate our approach on standard offline reinforcement learning benchmarks that involve challenging long-horizon planning. Furthermore, we show that, as a standalone module, LoMAP can be incorporated into the hierarchical diffusion planner, providing further performance enhancements.

Recent advances in autonomous vehicles have shifted development from rigid pipelines to end-to-end neural policies mapping raw sensor streams directly to control commands [1-3]. While these models offer streamlined architectures and strong benchmark performance, they raise critical deployment concerns. Their internal logic is opaque, complicating validation in safety-critical settings. They struggle to generalize to rare events like severe weather or infrastructure damage and lack formal guarantees on kinematic properties such as speed limits and lane-keeping. Further, they provide no natural interface for human oversight or explanation. These challenges motivate frameworks that combine deep network expressiveness with transparency, robustness, and provable safety. Meanwhile, Large Language Models (LLMs) and Vision Language Models (VLMs) have demonstrated human-level reasoning and visual grounding [4-6]. Recent works like VLM-SR (Shaped Rewards) [7], VLM-RM (Reward Models) [8], and RoboCLIP (Language-Conditioned Robot Learning via Contrastive Language-Image Pretraining) [9] inject semantic feedback into Reinforcement Learning (RL), but rely on static prompts unsuited to evolving road conditions and overlook vehicle dynamics.

Action-dependent individual policies, which incorporate both environmental states and the actions of other agents in decision-making, have emerged as a promising paradigm for achieving global optimality in multi-agent reinforcement learning (MARL). However, the existing literature often adopts auto-regressive action-dependent policies, where each agent's policy depends on the actions of all preceding agents. This formulation incurs substantial computational complexity as the number of agents increases, thereby limiting scalability. In this work, we consider a more generalized class of action-dependent policies, which do not necessarily follow the auto-regressive form. We propose to use the `action dependency graph (ADG)' to model the inter-agent action dependencies. Within the context of MARL problems structured by coordination graphs, we prove that an action-dependent policy with a sparse ADG can achieve global optimality, provided the ADG satisfies specific conditions specified by the coordination graph. Building on this theoretical foundation, we develop a tabular policy iteration algorithm with guaranteed global optimality. Furthermore, we integrate our framework into several SOTA algorithms and conduct experiments in complex environments. The empirical results affirm the robustness and applicability of our approach in more general scenarios, underscoring its potential for broader MARL challenges.

Plasticity, or the ability of an agent to adapt to new tasks, environments, or distributions, is crucial for continual learning. In this paper, we study the loss of plasticity in deep continual RL from the lens of churn: network output variability for out-of-batch data induced by mini-batch training. We demonstrate that (1) the loss of plasticity is accompanied by the exacerbation of churn due to the gradual rank decrease of the Neural Tangent Kernel (NTK) matrix; (2) reducing churn helps prevent rank collapse and adjusts the step size of regular RL gradients adaptively. Moreover, we introduce Continual Churn Approximated Reduction (C-CHAIN) and demonstrate it improves learning performance and outperforms baselines in a diverse range of continual learning environments on OpenAI Gym Control, ProcGen, DeepMind Control Suite, and MinAtar benchmarks.