Reinforcement Learning (RL) mathematically formulates decision-making with Markov Decision Process (MDP). With MDPs, researchers have achieved remarkable breakthroughs across various domains, including games, robotics, and language models. This paper seeks a new possibility, Natural Language Reinforcement Learning (NLRL), by extending traditional MDP to natural language-based representation space. Specifically, NLRL innovatively redefines RL principles, including task objectives, policy, value function, Bellman equation, and policy iteration, into their language counterparts. With recent advancements in large language models (LLMs), NLRL can be practically implemented to achieve RL-like policy and value improvement by either pure prompting or gradient-based training. Experiments over Maze, Breakthrough, and Tic-Tac-Toe games demonstrate the effectiveness, efficiency, and interpretability of the NLRL framework among diverse use cases. Our code will be released at https://github.com/waterhorse1/Natural-language-RL.

Recently in robotics, Vision-Language-Action (VLA) models have emerged as a transformative approach, enabling robots to execute complex tasks by integrating visual and linguistic inputs within an end-to-end learning framework. While VLA models offer significant capabilities, they also introduce new attack surfaces, making them vulnerable to adversarial attacks. With these vulnerabilities largely unexplored, this paper systematically quantifies the robustness of VLA-based robotic systems. Recognizing the unique demands of robotic execution, our attack objectives target the inherent spatial and functional characteristics of robotic systems. In particular, we introduce an untargeted position-aware attack objective that leverages spatial foundations to destabilize robotic actions, and a targeted attack objective that manipulates the robotic trajectory. Additionally, we design an adversarial patch generation approach that places a small, colorful patch within the camera's view, effectively executing the attack in both digital and physical environments. Our evaluation reveals a marked degradation in task success rates, with up to a 100\% reduction across a suite of simulated robotic tasks, highlighting critical security gaps in current VLA architectures. By unveiling these vulnerabilities and proposing actionable evaluation metrics, this work advances both the understanding and enhancement of safety for VLA-based robotic systems, underscoring the necessity for developing robust defense strategies prior to physical-world deployments.

Research on Reinforcement Learning (RL) approaches for discrete optimization problems has increased considerably, extending RL to an area classically dominated by Operations Research (OR). Vehicle routing problems are a good example of discrete optimization problems with high practical relevance where RL techniques have had considerable success. Despite these advances, open-source development frameworks remain scarce, hampering both the testing of algorithms and the ability to objectively compare results. This ultimately slows down progress in the field and limits the exchange of ideas between the RL and OR communities. Here we propose a library composed of multi-agent environments that simulates classic vehicle routing problems. The library, built on PyTorch, provides a flexible modular architecture design that allows easy customization and incorporation of new routing problems. It follows the Agent Environment Cycle ("AEC") games model and has an intuitive API, enabling rapid adoption and easy integration into existing reinforcement learning frameworks. The library allows for a straightforward use of classical OR benchmark instances in order to narrow the gap between the test beds for algorithm benchmarking used by the RL and OR communities. Additionally, we provide benchmark instance sets for each environment, as well as baseline RL models and training code.

Most reinforcement learning (RL) platforms use high-level programming languages, such as OpenAI Gymnasium using Python. These frameworks provide various API and benchmarks for testing RL algorithms in different domains, such as autonomous driving (AD) and robotics. These platforms often emphasise the design of RL algorithms and the training performance but neglect the correctness of models and reward functions, which can be crucial for the successful application of RL. This paper proposes using formal methods to model AD systems and demonstrates how model checking (MC) can be used in RL for AD. Most studies combining MC and RL focus on safety, such as safety shields. However, this paper shows different facets where MC can strengthen RL. First, an MC-based model pre-analysis can reveal bugs with respect to sensor accuracy and learning step size. This step serves as a preparation of RL, which saves time if bugs exist and deepens users' understanding of the target system. Second, reward automata can benefit the design of reward functions and greatly improve learning performance especially when the learning objectives are multiple. All these findings are supported by experiments.

Backscatter communication (BC) becomes a promising energy-efficient solution for future wireless sensor networks (WSNs). Unmanned aerial vehicles (UAVs) enable flexible data collection from remote backscatter devices (BDs), yet conventional UAVs rely on omni-directional fixed-position antennas (FPAs), limiting channel gain and prolonging data collection time. To address this issue, we consider equipping a UAV with a directional movable antenna (MA) with high directivity and flexibility. The MA enhances channel gain by precisely aiming its main lobe at each BD, focusing transmission power for efficient communication. Our goal is to minimize the total data collection time by jointly optimizing the UAV's trajectory and the MA's orientation. We develop a deep reinforcement learning (DRL)-based strategy using the azimuth angle and distance between the UAV and each BD to simplify the agent's observation space. To ensure stability during training, we adopt Soft Actor-Critic (SAC) algorithm that balances exploration with reward maximization for efficient and reliable learning. Simulation results demonstrate that our proposed MA-equipped UAV with SAC outperforms both FPA-equipped UAVs and other RL methods, achieving significant reductions in both data collection time and energy consumption.

Training agents that can coordinate zero-shot with humans is a key mission in multi-agent reinforcement learning (MARL). Current algorithms focus on training simulated human partner policies which are then used to train a Cooperator agent. The simulated human is produced either through behavior cloning over a dataset of human cooperation behavior, or by using MARL to create a population of simulated agents. However, these approaches often struggle to produce a Cooperator that can coordinate well with real humans, since the simulated humans fail to cover the diverse strategies and styles employed by people in the real world. We show \emph{learning a generative model of human partners} can effectively address this issue. Our model learns a latent variable representation of the human that can be regarded as encoding the human's unique strategy, intention, experience, or style. This generative model can be flexibly trained from any (human or neural policy) agent interaction data. By sampling from the latent space, we can use the generative model to produce different partners to train Cooperator agents. We evaluate our method -- \textbf{G}enerative \textbf{A}gent \textbf{M}odeling for \textbf{M}ulti-agent \textbf{A}daptation (GAMMA) -- on Overcooked, a challenging cooperative cooking game that has become a standard benchmark for zero-shot coordination. We conduct an evaluation with real human teammates, and the results show that GAMMA consistently improves performance, whether the generative model is trained on simulated populations or human datasets. Further, we propose a method for posterior sampling from the generative model that is biased towards the human data, enabling us to efficiently improve performance with only a small amount of expensive human interaction data.

Designing integrated circuits involves substantial complexity, posing challenges in revealing its potential applications - from custom digital cells to analog circuits. Despite extensive research over the past decades in building versatile and automated frameworks, there remains open room to explore more computationally efficient AI-based solutions. This paper introduces the graph composer GraCo, a novel method for synthesizing integrated circuits using reinforcement learning (RL). GraCo learns to construct a graph step-by-step, which is then converted into a netlist and simulated with SPICE. We demonstrate that GraCo is highly configurable, enabling the incorporation of prior design knowledge into the framework. We formalize how this prior knowledge can be utilized and, in particular, show that applying consistency checks enhances the efficiency of the sampling process. To evaluate its performance, we compare GraCo to a random baseline, which is known to perform well for smaller design space problems. We demonstrate that GraCo can discover circuits for tasks such as generating standard cells, including the inverter and the two-input NAND (NAND2) gate. Compared to a random baseline, GraCo requires 5x fewer sampling steps to design an inverter and successfully synthesizes a NAND2 gate that is 2.5x faster.

Estimation of the Average Treatment Effect (ATE) is a core problem in causal inference with strong connections to Off-Policy Evaluation in Reinforcement Learning. This paper considers the problem of adaptively selecting the treatment allocation probability in order to improve estimation of the ATE. The majority of prior work on adaptive ATE estimation focus on asymptotic guarantees, and in turn overlooks important practical considerations such as the difficulty of learning the optimal treatment allocation as well as hyper-parameter selection. Existing non-asymptotic methods are limited by poor empirical performance and exponential scaling of the Neyman regret with respect to problem parameters. In order to address these gaps, we propose and analyze the Clipped Second Moment Tracking (ClipSMT) algorithm, a variant of an existing algorithm with strong asymptotic optimality guarantees, and provide finite sample bounds on its Neyman regret. Our analysis shows that ClipSMT achieves exponential improvements in Neyman regret on two fronts: improving the dependence on $T$ from $O(\sqrt{T})$ to $O(\log T)$, as well as reducing the exponential dependence on problem parameters to a polynomial dependence. Finally, we conclude with simulations which show the marked improvement of ClipSMT over existing approaches.

Q-Learning is a fundamental off-policy reinforcement learning (RL) algorithm that has the objective of approximating action-value functions in order to learn optimal policies. Nonetheless, it has difficulties in reconciling bias with variance, particularly in the context of long-term rewards. This paper introduces Q($\Delta$)-Learning, an extension of TD($\Delta$) for the Q-Learning framework. TD($\Delta$) facilitates efficient learning over several time scales by breaking the Q($\Delta$)-function into distinct discount factors. This approach offers improved learning stability and scalability, especially for long-term tasks where discounting bias may impede convergence. Our methodology guarantees that each element of the Q($\Delta$)-function is acquired individually, facilitating expedited convergence on shorter time scales and enhancing the learning of extended time scales. We demonstrate through theoretical analysis and practical evaluations on standard benchmarks like Atari that Q($\Delta$)-Learning surpasses conventional Q-Learning and TD learning methods in both tabular and deep RL environments.

Recent advances in deep learning have brought attention to the possibility of creating advanced, general AI systems that outperform humans across many tasks. However, if these systems pursue unintended goals, there could be catastrophic consequences. A key prerequisite for AI systems pursuing unintended goals is whether they will behave in a coherent and goal-directed manner in the first place, optimizing for some unknown goal; there exists significant research trying to evaluate systems for said behaviors. However, the most rigorous definitions of goal-directedness we currently have are difficult to compute in real-world settings. Drawing upon this previous literature, we explore policy goal-directedness within reinforcement learning (RL) environments. In our findings, we propose a different family of definitions of the goal-directedness of a policy that analyze whether it is well-modeled as near-optimal for many (sparse) reward functions. We operationalize this preliminary definition of goal-directedness and test it in toy Markov decision process (MDP) environments. Furthermore, we explore how goal-directedness could be measured in frontier large-language models (LLMs). Our contribution is a definition of goal-directedness that is simpler and more easily computable in order to approach the question of whether AI systems could pursue dangerous goals. We recommend further exploration of measuring coherence and goal-directedness, based on our findings.