Efficient traffic control (TSC) is essential for urban mobility, but traditional systems struggle to handle the complexity of real-world traffic. Multi-agent Reinforcement Learning (MARL) offers adaptive solutions, but online MARL requires extensive interactions with the environment, making it costly and impractical. Offline MARL mitigates these challenges by using historical traffic data for training but faces significant difficulties with heterogeneous behavior policies in real-world datasets, where mixed-quality data complicates learning. We introduce OffLight, a novel offline MARL framework designed to handle heterogeneous behavior policies in TSC datasets. To improve learning efficiency, OffLight incorporates Importance Sampling (IS) to correct for distributional shifts and Return-Based Prioritized Sampling (RBPS) to focus on high-quality experiences. OffLight utilizes a Gaussian Mixture Variational Graph Autoencoder (GMM-VGAE) to capture the diverse distribution of behavior policies from local observations. Extensive experiments across real-world urban traffic scenarios show that OffLight outperforms existing offline RL methods, achieving up to a 7.8% reduction in average travel time and 11.2% decrease in queue length. Ablation studies confirm the effectiveness of OffLight's components in handling heterogeneous data and improving policy performance. These results highlight OffLight's scalability and potential to improve urban traffic management without the risks of online learning.

We introduce an Implicit Game-Theoretic MPC (IGT-MPC), a decentralized algorithm for two-agent motion planning that uses a learned value function that predicts the game-theoretic interaction outcomes as the terminal cost-to-go function in a model predictive control (MPC) framework, guiding agents to implicitly account for interactions with other agents and maximize their reward. This approach applies to competitive and cooperative multi-agent motion planning problems which we formulate as constrained dynamic games. Given a constrained dynamic game, we randomly sample initial conditions and solve for the generalized Nash equilibrium (GNE) to generate a dataset of GNE solutions, computing the reward outcome of each game-theoretic interaction from the GNE. The data is used to train a simple neural network to predict the reward outcome, which we use as the terminal cost-to-go function in an MPC scheme. We showcase emerging competitive and coordinated behaviors using IGT-MPC in scenarios such as two-vehicle head-to-head racing and un-signalized intersection navigation. IGT-MPC offers a novel method integrating machine learning and game-theoretic reasoning into model-based decentralized multi-agent motion planning.

Federated reinforcement learning (FedRL) enables multiple agents to collaboratively learn a policy without sharing their local trajectories collected during agent-environment interactions. However, in practice, the environments faced by different agents are often heterogeneous, leading to poor performance by the single policy learned by existing FedRL algorithms on individual agents. In this paper, we take a further step and introduce a \emph{personalized} FedRL framework (PFedRL) by taking advantage of possibly shared common structure among agents in heterogeneous environments. Specifically, we develop a class of PFedRL algorithms named PFedRL-Rep that learns (1) a shared feature representation collaboratively among all agents, and (2) an agent-specific weight vector personalized to its local environment. We analyze the convergence of PFedTD-Rep, a particular instance of the framework with temporal difference (TD) learning and linear representations. To the best of our knowledge, we are the first to prove a linear convergence speedup with respect to the number of agents in the PFedRL setting. To achieve this, we show that PFedTD-Rep is an example of the federated two-timescale stochastic approximation with Markovian noise. Experimental results demonstrate that PFedTD-Rep, along with an extension to the control setting based on deep Q-networks (DQN), not only improve learning in heterogeneous settings, but also provide better generalization to new environments.

Planning with partial observation is a central challenge in embodied AI. A majority of prior works have tackled this challenge by developing agents that physically explore their environment to update their beliefs about the world state. In contrast, humans can $\textit{imagine}$ unseen parts of the world through a mental exploration and $\textit{revise}$ their beliefs with imagined observations. Such updated beliefs can allow them to make more informed decisions, without necessitating the physical exploration of the world at all times. To achieve this human-like ability, we introduce the $\textit{Generative World Explorer (Genex)}$, an egocentric world exploration framework that allows an agent to mentally explore a large-scale 3D world (e.g., urban scenes) and acquire imagined observations to update its belief. This updated belief will then help the agent to make a more informed decision at the current step. To train $\textit{Genex}$, we create a synthetic urban scene dataset, Genex-DB. Our experimental results demonstrate that (1) $\textit{Genex}$ can generate high-quality and consistent observations during long-horizon exploration of a large virtual physical world and (2) the beliefs updated with the generated observations can inform an existing decision-making model (e.g., an LLM agent) to make better plans.

Medical Decision-Making (MDM) is a multi-faceted process that requires clinicians to assess complex multi-modal patient data patient, often collaboratively. Large Language Models (LLMs) promise to streamline this process by synthesizing vast medical knowledge and multi-modal health data. However, single-agent are often ill-suited for nuanced medical contexts requiring adaptable, collaborative problem-solving. Our MDAgents addresses this need by dynamically assigning collaboration structures to LLMs based on task complexity, mimicking real-world clinical collaboration and decision-making. This framework improves diagnostic accuracy and supports adaptive responses in complex, real-world medical scenarios, making it a valuable tool for clinicians in various healthcare settings, and at the same time, being more efficient in terms of computing cost than static multi-agent decision making methods.

Multi-agent network systems have found applications in various societal infrastructures, such as power systems, traffic networks, and smart cities [McArthur et al., 2007, Burmeister et al., 1997, Roscia et al., 2013]. One particularly important class of such problems is the cooperative multi-agent network MDP setting, where agents are embedded in a graph, and each agent has its own local state [Qu et al., 2020b]. In network MDPs, the local state transition probabilities and rewards only depend on the states and actions of the agent's direct neighbors in the graph. Such a property has been observed in a great variety of cooperative network control problems, ranging from thermal control of multizone buildings [Zhang et al., 2016], wireless access control [Zocca, 2019] to phase synchronization in electrical grids [Blaabjerg et al., 2006], where agents typically only need to act and learn based on information within a local neighborhood due to constraints on the information and communication infrastructure.

Matching markets have become increasingly relevant in a variety of modern applications, including but not limited to school admissions, organ transplantation, and job matching. Traditionally, these markets were studied under the assumption that the demand side agents (aka players or agents) and the supply side agents (aka arms) have fixed, known preferences, allowing for stable matching via the deferred acceptance algorithm like the Gale-Shapley algorithm introduced in Gale and Shapley (1962). However, in applications like crowdsourcing, online labor markets, and finance, the preferences are not given to the agents, and they must learn it over time by interacting with the environment. Modeling the matching market as multi-agent, multi-armed competitive bandits, there has been extensive work on various aspects, including coordinated centralized matching, decentralized matching, and game-theoretic analysis (see Liu et al. (2020); Basu et al. (2021); Sankararaman et al. (2021); Etesami and Srikant (2024)). In this paper, we study a large and dynamic matching market where the number of arms K is large and often exceeds the number of agents N( K).

In decentralized multi-agent reinforcement learning, agents learning in isolation can lead to relative over-generalization (RO), where optimal joint actions are undervalued in favor of suboptimal ones. This hinders effective coordination in cooperative tasks, as agents tend to choose actions that are individually rational but collectively suboptimal. To address this issue, we introduce MaxMax Q-Learning (MMQ), which employs an iterative process of sampling and evaluating potential next states, selecting those with maximal Q-values for learning. This approach refines approximations of ideal state transitions, aligning more closely with the optimal joint policy of collaborating agents. We provide theoretical analysis supporting MMQ's potential and present empirical evaluations across various environments susceptible to RO. Our results demonstrate that MMQ frequently outperforms existing baselines, exhibiting enhanced convergence and sample efficiency.

Artificial Intelligence is rapidly embedding itself within militaries, economies, and societies, reshaping their very foundations. Given the depth and breadth of its consequences, it has never been more pressing to understand how to ensure that AI systems are safe, ethical, and have a positive societal impact. This book aims to provide a comprehensive approach to understanding AI risk. Our primary goals include consolidating fragmented knowledge on AI risk, increasing the precision of core ideas, and reducing barriers to entry by making content simpler and more comprehensible. The book has been designed to be accessible to readers from diverse backgrounds. You do not need to have studied AI, philosophy, or other such topics. The content is skimmable and somewhat modular, so that you can choose which chapters to read. We introduce mathematical formulas in a few places to specify claims more precisely, but readers should be able to understand the main points without these.

Autonomous multi-agent systems such as hospital robots and package delivery drones often operate in highly uncertain environments and are expected to achieve complex temporal task objectives while ensuring safety. While learning-based methods such as reinforcement learning are popular methods to train single and multi-agent autonomous systems under user-specified and state-based reward functions, applying these methods to satisfy trajectory-level task objectives is a challenging problem. Our first contribution is the use of weighted automata to specify trajectory-level objectives, such that, maximal paths induced in the weighted automaton correspond to desired trajectory-level behaviors. We show how weighted automata-based specifications go beyond timeliness properties focused on deadlines to performance properties such as expeditiousness. Our second contribution is the use of evolutionary game theory (EGT) principles to train homogeneous multi-agent teams targeting homogeneous task objectives. We show how shared experiences of agents and EGT-based policy updates allow us to outperform state-of-the-art reinforcement learning (RL) methods in minimizing path length by nearly 30\% in large spaces. We also show that our algorithm is computationally faster than deep RL methods by at least an order of magnitude. Additionally our results indicate that it scales better with an increase in the number of agents as compared to other methods.