Multi-agent systems, where specialized agents collaborate to solve a shared task hold great potential, from increased modularity to simulating complex environments. However, they also have a major caveat -- a single agent can cause the entire system to fail. Consider a simple game where the knowledge to solve the task is distributed between agents, which share information in a communication channel. At each round, any of the agents can terminate the game and make the final prediction, even if they are uncertain about the outcome of their action. Detection of such rogue agents $\textit{before they act}$ may prevent the system's failure. In this work, we propose to $\textit{monitor}$ agents during action prediction and $\textit{intervene}$ when a future error is likely to occur. To test our approach, we introduce WhoDunitEnv, a multi-agent collaboration environment that allows modular control over task complexity and communication structure. Experiments on two variants of WhoDunitEnv and the GovSim environment for resource sustainability show that our approach leads to substantial performance gains up to 17.4% and 20%, respectively. Moreover, a thorough analysis shows that our monitors successfully identify critical points of agent confusion and our interventions effectively stop agent errors from propagating.

Optimization is instrumental for improving operations of large-scale socio-technical infrastructures of Smart Cities, for instance, energy and traffic systems. In particular, understanding the performance of multi-agent discrete-choice combinatorial optimization under distributed adversary attacks is a compelling and underexplored problem, since multi-agent systems exhibit a large number of remote control variables that can influence in an unprecedented way the cost-effectiveness of distributed optimization heuristics. This paper unravels for the first time the trajectories of distributed optimization from resilience to vulnerability, and finally to collapse under varying adversary influence. Using real-world data to emulate over 28 billion multi-agent optimization scenarios, we exhaustively assess how the number of agents with different adversarial severity and network positioning influences optimization performance, including the influence on Pareto optimal points. With this novel large-scale dataset, made openly available as a benchmark, we disentangle how optimization remains resilient to adversaries and which adversary conditions are required to make optimization vulnerable or collapsed. These new findings can provide new insights for designing self-healing strategies for fault-tolerance and fault-correction in adversarial distributed optimization that have been missing so far.

While much work on web agents emphasizes the promise of autonomously performing tasks on behalf of users, in reality, agents often fall short on complex tasks in real-world contexts and modeling user preference. This presents an opportunity for humans to collaborate with the agent and leverage the agent's capabilities effectively. We propose CowPilot, a framework supporting autonomous as well as human-agent collaborative web navigation, and evaluation across task success and task efficiency. CowPilot reduces the number of steps humans need to perform by allowing agents to propose next steps, while users are able to pause, reject, or take alternative actions. During execution, users can interleave their actions with the agent by overriding suggestions or resuming agent control when needed. We conducted case studies on five common websites and found that the human-agent collaborative mode achieves the highest success rate of 95% while requiring humans to perform only 15.2% of the total steps. Even with human interventions during task execution, the agent successfully drives up to half of task success on its own. CowPilot can serve as a useful tool for data collection and agent evaluation across websites, which we believe will enable research in how users and agents can work together. Video demonstrations are available at https://oaishi.github.io/cowpilot.html

Under what conditions can capable AI agents efficiently align their actions with human preferences? More specifically, when they are proficient enough to collaborate with us, how long does coordination take, and when is it computationally feasible? These foundational questions of AI alignment help define what makes an AI agent ``sufficiently safe'' and valuable to humans. Since such generally capable systems do not yet exist, a theoretical analysis is needed to establish when guarantees hold -- and what they even are. We introduce a game-theoretic framework that generalizes prior alignment approaches with fewer assumptions, allowing us to analyze the computational complexity of alignment across $M$ objectives and $N$ agents, providing both upper and lower bounds. Unlike previous work, which often assumes common priors, idealized communication, or implicit tractability, our framework formally characterizes the difficulty of alignment under minimal assumptions. Our main result shows that even when agents are fully rational and computationally \emph{unbounded}, alignment can be achieved with high probability in time \emph{linear} in the task space size. Therefore, in real-world settings, where task spaces are often \emph{exponential} in input length, this remains impractical. More strikingly, our lower bound demonstrates that alignment is \emph{impossible} to speed up when scaling to exponentially many tasks or agents, highlighting a fundamental computational barrier to scalable alignment. Relaxing these idealized assumptions, we study \emph{computationally bounded} agents with noisy messages (representing obfuscated intent), showing that while alignment can still succeed with high probability, it incurs additional \emph{exponential} slowdowns in the task space size, number of agents, and number of tasks. We conclude by identifying conditions that make alignment more feasible.

We study the problem of adapting to a known sub-rational opponent during online play while remaining robust to rational opponents. We focus on large imperfect-information (zero-sum) games, which makes it impossible to inspect the whole game tree at once and necessitates the use of depth-limited search. However, all existing methods assume rational play beyond the depth-limit, which only allows them to adapt a very limited portion of the opponent's behaviour. We propose an algorithm Adapting Beyond Depth-limit (ABD) that uses a strategy-portfolio approach - which we refer to as matrix-valued states - for depth-limited search. This allows the algorithm to fully utilise all information about the opponent model, making it the first robust-adaptation method to be able to do so in large imperfect-information games. As an additional benefit, the use of matrix-valued states makes the algorithm simpler than traditional methods based on optimal value functions. Our experimental results in poker and battleship show that ABD yields more than a twofold increase in utility when facing opponents who make mistakes beyond the depth limit and also delivers significant improvements in utility and safety against randomly generated opponents.

Recent advances in large-scale machine learning have produced high-capacity foundation models capable of adapting to a broad array of downstream tasks. While such models hold great promise for robotics, the prevailing paradigm still portrays robots as single, autonomous decision-makers, performing tasks like manipulation and navigation, with limited human involvement. However, a large class of real-world robotic systems, including wearable robotics (e.g., prostheses, orthoses, exoskeletons), teleoperation, and neural interfaces, are semiautonomous, and require ongoing interactive coordination with human partners, challenging single-agent assumptions. In this position paper, we argue that robot foundation models must evolve to an interactive multi-agent perspective in order to handle the complexities of real-time human-robot co-adaptation. We propose a generalizable, neuroscience-inspired architecture encompassing four modules: (1) a multimodal sensing module informed by sensorimotor integration principles, (2) an ad-hoc teamwork model reminiscent of joint-action frameworks in cognitive science, (3) a predictive world belief model grounded in internal model theories of motor control, and (4) a memory/feedback mechanism that echoes concepts of Hebbian and reinforcement-based plasticity. Although illustrated through the lens of cyborg systems, where wearable devices and human physiology are inseparably intertwined, the proposed framework is broadly applicable to robots operating in semi-autonomous or interactive contexts. By moving beyond single-agent designs, our position emphasizes how foundation models in robotics can achieve a more robust, personalized, and anticipatory level of performance.

Communicating in natural language is a powerful tool in multi-agent settings, as it enables independent agents to share information in partially observable settings and allows zero-shot coordination with humans. However, most prior works are limited as they either rely on training with large amounts of human demonstrations or lack the ability to generate natural and useful communication strategies. In this work, we train language models to have productive discussions about their environment in natural language without any human demonstrations. We decompose the communication problem into listening and speaking. Our key idea is to leverage the agent's goal to predict useful information about the world as a dense reward signal that guides communication. Specifically, we improve a model's listening skills by training them to predict information about the environment based on discussions, and we simultaneously improve a model's speaking skills with multi-agent reinforcement learning by rewarding messages based on their influence on other agents. To investigate the role and necessity of communication in complex social settings, we study an embodied social deduction game based on Among Us, where the key question to answer is the identity of an adversarial imposter. We analyze emergent behaviors due to our technique, such as accusing suspects and providing evidence, and find that it enables strong discussions, doubling the win rates compared to standard RL. We release our code and models at https://socialdeductionllm.github.io/

This paper introduces a new mission planning algorithm for autonomous robots that enables the reward-based selection of an optimal waypoint sequence from a predefined set. The algorithm computes a feasible trajectory and corresponding control inputs for a robot to navigate between waypoints while avoiding obstacles, maximizing the total reward, and adhering to constraints on state, input and its derivatives, mission time window, and maximum distance. This also solves a generalized prize-collecting traveling salesman problem. The proposed algorithm employs a new genetic algorithm that evolves solution candidates toward the optimal solution based on a fitness function and crossover. During fitness evaluation, a penalty method enforces constraints, and the differential flatness property with clothoid curves efficiently penalizes infeasible trajectories. The Euler spiral method showed promising results for trajectory parameterization compared to minimum snap and jerk polynomials. Due to the discrete exploration space, crossover is performed using a dynamic time-warping-based method and extended convex combination with projection. A mutation step enhances exploration. Results demonstrate the algorithm's ability to find the optimal waypoint sequence, fulfill constraints, avoid infeasible waypoints, and prioritize high-reward ones. Simulations and experiments with a ground vehicle, quadrotor, and quadruped are presented, complemented by benchmarking and a time-complexity analysis.

Under-resourced or rural hospitals have limited access to medical specialists and healthcare professionals, which can negatively impact patient outcomes in sepsis. To address this gap, we developed the MATEC (Multi-AI Agent Team Care) framework, which integrates a team of specialized AI agents for sepsis care. The sepsis AI agent team includes five doctor agents, four health professional agents, and a risk prediction model agent, with an additional 33 doctor agents available for consultations. Ten attending physicians at a teaching hospital evaluated this framework, spending approximately 40 minutes on the web-based MATEC application and participating in the 5-point Likert scale survey (rated from 1-unfavorable to 5-favorable). The physicians found the MATEC framework very useful (Median=4, P=0.01), and very accurate (Median=4, P<0.01). This pilot study demonstrates that a Multi-AI Agent Team Care framework (MATEC) can potentially be useful in assisting medical professionals, particularly in under-resourced hospital settings.

Weaknesses: The experiments have been extensive, however I have following three crucial questions to better understand the performance boost arising from the overall architecture: 1. Improvement arising from interaction module or motion module? Taking Social LSTM [1] to be an interaction-based baseline, the proposed architecture has two different components: the interaction and motion modules. Is the boost coming from the interaction module which is FQA in comparison to Social Pooling [1]? Or is it the new motion module? An ablation study showing the performance while keeping the motion module the same as the baseline will help answer this question. The authors use the term Fuzzy to describe continuous-valued decisions over their discrete-valued boolean counterparts.