Multi-agent self-organizing systems (MASOS) exhibit key characteristics including scalability, adaptability, flexibility, and robustness, which have contributed to their extensive application across various fields. However, the self-organizing nature of MASOS also introduces elements of unpredictability in their emergent behaviors. This paper focuses on the emergence of dependency hierarchies during task execution, aiming to understand how such hierarchies arise from agents' collective pursuit of the joint objective, how they evolve dynamically, and what factors govern their development. To investigate this phenomenon, multi-agent reinforcement learning (MARL) is employed to train MASOS for a collaborative box-pushing task. By calculating the gradients of each agent's actions in relation to the states of other agents, the inter-agent dependencies are quantified, and the emergence of hierarchies is analyzed through the aggregation of these dependencies. Our results demonstrate that hierarchies emerge dynamically as agents work towards a joint objective, with these hierarchies evolving in response to changing task requirements. Notably, these dependency hierarchies emerge organically in response to the shared objective, rather than being a consequence of pre-configured rules or parameters that can be fine-tuned to achieve specific results. Furthermore, the emergence of hierarchies is influenced by the task environment and network initialization conditions. Additionally, hierarchies in MASOS emerge from the dynamic interplay between agents' "Talent" and "Effort" within the "Environment." "Talent" determines an agent's initial influence on collective decision-making, while continuous "Effort" within the "Environment" enables agents to shift their roles and positions within the system.

Collective behaviors such as swarming and flocking emerge from simple, decentralized interactions in biological systems. Existing models, such as Vicsek and Cucker-Smale, lack collision avoidance, whereas the Olfati-Saber model imposes rigid formations, limiting their applicability in swarm robotics. To address these limitations, this paper proposes a minimal yet expressive model that governs agent dynamics using relative positions, velocities, and local density, modulated by two tunable parameters: the spatial offset and kinetic offset. The model achieves spatially flexible, collision-free behaviors that reflect naturalistic group dynamics. Furthermore, we extend the framework to cognitive autonomous systems, enabling energy-aware phase transitions between swarming and flocking through adaptive control parameter tuning. This cognitively inspired approach offers a robust foundation for real-world applications in multi-robot systems, particularly autonomous aerial swarms.

Hegselmann--Krause models are localized, distributed averaging dynamics on spatial data. A key aspect of these dynamics is that they lead to cluster formation, which has important applications in geographic information systems, dynamic clustering algorithms, opinion dynamics, and social networks. For these models, the key questions are whether a fixed point exists and, if so, characterizing it. In this work, we establish new results towards the "2R-Conjecture" for the Hegselmann--Krause model, for which no meaningful progress, or even any precise statement, has been made since its introduction in 2007. This conjecture relates to the structure of the fixed point when there are a large number of agents per unit space. We provide, among other results, a proof in expectation and a statement of a stronger result that is supported by simulation. The key methodological contribution is to consider the dynamics as an infinite-dimensional problem on the space of point processes, rather than on finitely many points. This enables us to leverage stationarity, shift invariance, and certain other symmetries to obtain the results. These techniques do not have finite-dimensional analogs.

In multi-agent systems, the safe and reliable execution of tasks often depends on agents correctly coordinating their actions. However, in real-world deployments, failures of computational components are inevitable, presenting a critical challenge: ensuring that multi-agent reinforcement learning (MARL) policies remain effective even when some agents malfunction. We propose the Multi-Agent Robust Training Algorithm (MARTA), a plug-and-play framework for training MARL agents to be resilient to potentially severe faults. MARTA operates in cooperative multi-agent settings where agents may lose the ability to execute their intended actions. It learns to identify failure scenarios that are especially detrimental to system performance and equips agents with strategies to mitigate their impact. At the heart of MARTA is a novel adversarial Markov game in which an adversary -- modelled via \emph{Markov switching controls} -- learns to disable agents in high-risk state regions, while the remaining agents are trained to \emph{jointly} best-respond to such targeted malfunctions. To ensure practicality, MARTA enforces a malfunction budget, constraining the adversary to a fixed number of failures and learning robust policies accordingly. We provide theoretical guarantees that MARTA converges to a Markov perfect equilibrium, ensuring agents optimally counteract worst-case faults. Empirically, we show that MARTA achieves state-of-the-art fault-tolerant performance across benchmark environments, including Multi-Agent Particle World and Level-Based Foraging.

As multi-agent systems (MAS) become increasingly prevalent in autonomous systems, distributed control, and edge intelligence, efficient communication under resource constraints has emerged as a critical challenge. Traditional communication paradigms often emphasize message fidelity or bandwidth optimization, overlooking the task relevance of the exchanged information. In contrast, goal-oriented communication prioritizes the importance of information with respect to the agents' shared objectives. This review provides a comprehensive survey of goal-oriented communication in MAS, bridging perspectives from information theory, communication theory, and machine learning. We examine foundational concepts alongside learning-based approaches and emergent protocols. Special attention is given to coordination under communication constraints, as well as applications in domains such as swarm robotics, federated learning, and edge computing. The paper concludes with a discussion of open challenges and future research directions at the intersection of communication theory, machine learning, and multi-agent decision making.

In this paper, we investigate: ``How can a group of initially co-located mobile agents perpetually explore an unknown graph, when one stationary node occasionally behaves maliciously, under an adversary's control?'' We call this node a ``Byzantine black hole (BBH)'' and at any given round it may choose to destroy all visiting agents, or none. This subtle power can drastically undermine classical exploration strategies designed for an always active black hole. We study this perpetual exploration problem in the presence of at most one BBH, without initial knowledge of the network size. Since the underlying graph may be 1-connected, perpetual exploration of the entire graph may be infeasible. We thus define two variants: \pbmPerpExpl\ and \pbmPerpExplHome. In the former, the agents are tasked to perform perpetual exploration of at least one component, obtained after the exclusion of the BBH. In the latter, the agents are tasked to perform perpetual exploration of the component which contains the \emph{home} node, where agents are initially co-located. Naturally, \pbmPerpExplHome\ is a special case of \pbmPerpExpl. Agents operate under a synchronous scheduler and communicate in a face-to-face model. Our goal is to determine the minimum number of agents necessary and sufficient to solve these problems. In acyclic networks, we obtain optimal algorithms that solve \pbmPerpExpl\ with $4$ agents, and \pbmPerpExplHome\ with $6$ agents in trees. The lower bounds hold even in path graphs. In general graphs, we give a non-trivial lower bound of $2Δ-1$ agents for \pbmPerpExpl, and an upper bound of $3Δ+3$ agents for \pbmPerpExplHome. To our knowledge, this is the first study of a black-hole variant in arbitrary networks without initial topological knowledge.

Language is a ubiquitous tool that is foundational to reasoning and collaboration, ranging from everyday interactions to sophisticated problem-solving tasks. The establishment of a common language can serve as a powerful asset in ensuring clear communication and understanding amongst agents, facilitating desired coordination and strategies. In this work, we extend the capabilities of large language models (LLMs) by integrating them with advancements in multi-agent decision-making algorithms. We propose a systematic framework for the design of multi-agentic large language models (LLMs), focusing on key integration practices. These include advanced prompt engineering techniques, the development of effective memory architectures, multi-modal information processing, and alignment strategies through fine-tuning algorithms. We evaluate these design choices through extensive ablation studies on classic game settings with significant underlying social dilemmas and game-theoretic considerations.

The Agentic Service Ecosystem consists of heterogeneous autonomous agents (e.g., intelligent machines, humans, and human-machine hybrid systems) that interact through resource exchange and service co-creation. These agents, with distinct behaviors and motivations, exhibit autonomous perception, reasoning, and action capabilities, which increase system complexity and make traditional linear analysis methods inadequate. Swarm intelligence, characterized by decentralization, self-organization, emergence, and dynamic adaptability, offers a novel theoretical lens and methodology for understanding and optimizing such ecosystems. However, current research, owing to fragmented perspectives and cross-ecosystem differences, fails to comprehensively capture the complexity of swarm-intelligence emergence in agentic contexts. The lack of a unified methodology further limits the depth and systematic treatment of the research. This paper proposes a framework for analyzing the emergence of swarm intelligence in Agentic Service Ecosystems, with three steps: measurement, analysis, and optimization, to reveal the cyclical mechanisms and quantitative criteria that foster emergence. By reviewing existing technologies, the paper analyzes their strengths and limitations, identifies unresolved challenges, and shows how this framework provides both theoretical support and actionable methods for real-world applications.

Cooperative multi-agent reinforcement learning (MARL) aims to coordinate multiple agents to achieve a common goal. A key challenge in MARL is credit assignment, which involves assessing each agent's contribution to the shared reward. Given the diversity of tasks, agents may perform different types of coordination, with rewards attributed to diverse and often overlapping agent subsets. In this work, we formalize the credit assignment level as the number of agents cooperating to obtain a reward, and address scenarios with multiple coexisting levels. We introduce a multi-level advantage formulation that performs explicit counterfactual reasoning to infer credits across distinct levels. Our method, Multi-level Advantage Credit Assignment (MACA), captures agent contributions at multiple levels by integrating advantage functions that reason about individual, joint, and correlated actions. Utilizing an attention-based framework, MACA identifies correlated agent relationships and constructs multi-level advantages to guide policy learning. Comprehensive experiments on challenging Starcraft v1\&v2 tasks demonstrate MACA's superior performance, underscoring its efficacy in complex credit assignment scenarios.

Generative models, especially diffusion and flow-based models, have been promising in offline multi-agent reinforcement learning. However, integrating powerful generative models into this framework poses unique challenges. In particular, diffusion and flow-based policies suffer from low sampling efficiency due to their iterative generation processes, making them impractical in time-sensitive or resource-constrained settings. To tackle these difficulties, we propose OM2P (Offline Multi-Agent Mean-Flow Policy), a novel offline MARL algorithm to achieve efficient one-step action sampling. To address the misalignment between generative objectives and reward maximization, we introduce a reward-aware optimization scheme that integrates a carefully-designed mean-flow matching loss with Q-function supervision. Additionally, we design a generalized timestep distribution and a derivative-free estimation strategy to reduce memory overhead and improve training stability. Empirical evaluations on Multi-Agent Particle and MuJoCo benchmarks demonstrate that OM2P achieves superior performance, with up to a 3.8x reduction in GPU memory usage and up to a 10.8x speed-up in training time. Our approach represents the first to successfully integrate mean-flow model into offline MARL, paving the way for practical and scalable generative policies in cooperative multi-agent settings.