The remarkable growth in large language model (LLM) capabilities has spurred exploration into multi-agent systems, with debate frameworks emerging as a promising avenue for enhanced problem-solving. These multi-agent debate (MAD) approaches, where agents collaboratively present, critique, and refine arguments, potentially offer improved reasoning, robustness, and diverse perspectives over monolithic models. Despite prior studies leveraging MAD, a systematic understanding of its effectiveness compared to self-agent methods, particularly under varying conditions, remains elusive. This paper seeks to fill this gap by conceptualizing MAD as a test-time computational scaling technique, distinguished by collaborative refinement and diverse exploration capabilities. We conduct a comprehensive empirical investigation comparing MAD with strong self-agent test-time scaling baselines on mathematical reasoning and safety-related tasks. Our study systematically examines the influence of task difficulty, model scale, and agent diversity on MAD's performance. Key findings reveal that, for mathematical reasoning, MAD offers limited advantages over self-agent scaling but becomes more effective with increased problem difficulty and decreased model capability, while agent diversity shows little benefit. Conversely, for safety tasks, MAD's collaborative refinement can increase vulnerability, but incorporating diverse agent configurations facilitates a gradual reduction in attack success through the collaborative refinement process. We believe our findings provide critical guidance for the future development of more effective and strategically deployed MAD systems.

Stochastic linear bandits are a fundamental model in sequential decision-making, where in each round, an agent selects a vector-valued action and receives a stochastic reward with expectation equal to the inner product between the action and an unknown parameter vector [14]. This setting has been extensively studied in the single-agent case, where the goal is to maximize cumulative reward over time through efficient exploration and exploitation [19, 12, 1]. In this paper, we consider a generalized multi-agent setting, motivated by many practical scenarios such as distributed decision systems [13, 24], personalized adaptive interfaces [16, 11], and multi-device learning [4]--where in each round, a set of m agents must each select an action and receive feedback from a stochastic, potentially heterogeneous reward function. A natural baseline approach is to run a standard single-agent linear bandit algorithm independently for each agent. While straightforward, this approach entirely ignores potential similarities across agents. When the agents' environments are similar, such independent learning leads to redundant exploration and a suboptimal use of data. In contrast, our goal is to develop collaborative learning algorithms that adaptively share information across agents--without knowing in advance how similar or different they are. This raises a fundamental question: How can collaboration reduce regret under unknown heterogeneity? To address the question, we model heterogeneity using a hierarchical/empirical Bayesian framework, where each agent's unknown parameter vector is drawn from an unknown population distribution.

--The RoboRacer (F1TENTH) platform has emerged as a leading testbed for advancing autonomous driving research, offering a scalable, cost-effective, and community-driven environment for experimentation. This paper presents a comprehensive survey of the platform, analyzing its modular hardware and software architecture, diverse research applications, and role in autonomous systems education. We examine critical aspects such as bridging the simulation-to-reality (Sim2Real) gap, integration with simulation environments, and the availability of standardized datasets and benchmarks. Furthermore, the survey highlights advancements in perception, planning, and control algorithms, as well as insights from global competitions and collaborative research efforts. The findings underscore the platform's significance in driving forward developments in autonomous racing and robotics.

-- Risk quantification is a critical component of safe autonomous driving, however, constrained by the limited perception range and occlusion of single-vehicle systems in complex and dense scenarios. V ehicle-to-everything (V2X) paradigm has been a promising solution to sharing complementary perception information, nevertheless, how to ensure the risk interpretability while understanding multi-agent interaction with V2X remains an open question. In this paper, we introduce the first V2X-enabled risk quantification pipeline, CooperRisk, to fuse perception information from multiple agents and quantify the scenario driving risk in future multiple timestamps. The risk is represented as a scenario risk map to ensure interpretability based on risk severity and exposure, and the multi-agent interaction is captured by the learning-based cooperative prediction model. It aims to ensure scene-consistent future behaviors of multiple agents and avoid conflicting predictions that could lead to overly conservative risk quantification and cause the ego vehicle to become overly hesitant to drive. Then, the temporal risk maps could serve to guide a model predictive control planner . We evaluate the CooperRisk pipeline in a real-world V2X dataset V2XPnP, and the experiments demonstrate its superior performance in risk quantification, showing a 44.35% decrease in conflict rate between the ego vehicle and background traffic participants. I. INTRODUCTION Connected autonomous vehicle (CA V) technology has been regarded as one of the most promising solutions for safe transportation [1]-[3].

Cooperative multi-agent systems often face tasks that require coordinated actions under uncertainty. While multi-armed bandit (MAB) problems provide a powerful framework for decentralized learning, most prior work assumes individually attainable rewards. We address the challenging setting where rewards are threshold-activated: an arm yields a payoff only when a minimum number of agents pull it simultaneously, with this threshold unknown in advance. Complicating matters further, some arms are decoys - requiring coordination to activate but yielding no reward - introducing a new challenge of wasted joint exploration. We introduce Threshold-Coop-UCB (T-Coop-UCB), a decentralized algorithm that enables agents to jointly learn activation thresholds and reward distributions, forming effective coalitions without centralized control. Empirical results show that T-Coop-UCB consistently outperforms baseline methods in cumulative reward, regret, and coordination metrics, achieving near-Oracle performance. Our findings underscore the importance of joint threshold learning and decoy avoidance for scalable, decentralized cooperation in complex multi-agent

We propose co-creative learning as a novel paradigm where humans and AI, i.e., biological and artificial agents, mutually integrate their partial perceptual information and knowledge to construct shared external representations, a process we interpret as symbol emergence. Unlike traditional AI teaching based on unilateral knowledge transfer, this addresses the challenge of integrating information from inherently different modalities. We empirically test this framework using a human-AI interaction model based on the Metropolis-Hastings naming game (MHNG), a decentralized Bayesian inference mechanism. In an online experiment, 69 participants played a joint attention naming game (JA-NG) with one of three computer agent types (MH-based, always-accept, or always-reject) under partial observability. Results show that human-AI pairs with an MH-based agent significantly improved categorization accuracy through interaction and achieved stronger convergence toward a shared sign system. Furthermore, human acceptance behavior aligned closely with the MH-derived acceptance probability. These findings provide the first empirical evidence for co-creative learning emerging in human-AI dyads via MHNG-based interaction. This suggests a promising path toward symbiotic AI systems that learn with humans, rather than from them, by dynamically aligning perceptual experiences, opening a new venue for symbiotic AI alignment.

Large Language Models (LLMs) have shown promise as decision-makers in dynamic settings, but their stateless nature necessitates creating a natural language representation of history. We present a unifying framework for systematically constructing natural language "state" representations for prompting LLM agents in repeated multi-agent games. Previous work on games with LLM agents has taken an ad hoc approach to encoding game history, which not only obscures the impact of state representation on agents' behavior, but also limits comparability between studies. Our framework addresses these gaps by characterizing methods of state representation along three axes: action informativeness (i.e., the extent to which the state representation captures actions played); reward informativeness (i.e., the extent to which the state representation describes rewards obtained); and prompting style (or natural language compression, i.e., the extent to which the full text history is summarized). We apply this framework to a dynamic selfish routing game, chosen because it admits a simple equilibrium both in theory and in human subject experiments \cite{rapoport_choice_2009}. Despite the game's relative simplicity, we find that there are key dependencies of LLM agent behavior on the natural language state representation. In particular, we observe that representations which provide agents with (1) summarized, rather than complete, natural language representations of past history; (2) information about regrets, rather than raw payoffs; and (3) limited information about others' actions lead to behavior that more closely matches game theoretic equilibrium predictions, and with more stable game play by the agents. By contrast, other representations can exhibit either large deviations from equilibrium, higher variation in dynamic game play over time, or both.

--The computation demands from the maritime Internet of Things (MIoT) increase rapidly in recent years, and the unmanned aerial vehicles (UA Vs) and vessels based multi-access edge computing (MEC) can fulfill these MIoT requirements. In this paper, we focus on the maritime computation offloading and resource allocation through the cooperation of UA Vs and vessels, with consideration of uncertain tasks. Specifically, we propose a cooperative MEC framework for computation offloading and resource allocation, including MIoT devices, UA Vs and vessels. Then, we formulate the optimization problem to minimize the total execution time. As for the uncertain MIoT tasks, we leverage Lyapunov optimization to tackle the unpredictable task arrivals and varying computational resource availability. By converting the long-term constraints into short-term constraints, we obtain a set of small-scale optimization problems. Moreover, a heterogeneous-agent soft actor-critic is proposed to sequentially update various neural networks and effectively solve the MG problem. Finally, simulations are conducted to verify the effectiveness in addressing computational offloading and resource allocation. Then, the maritime Internet of Things (MIoT) employs sensors and wireless networks to collect, transmit, analyze data, and enhance the intelligence of maritime management. Jiahao Y ou, Chao Dong, and Qihui Wu are with the College of Electronic and Information Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China (e-mail: yjiahao@nuaa.edu.cn, Ziye Jia is with the College of Electronic and Information Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China, and also with the National Mobile Communications Research Laboratory, Southeast University, Nanjing, Jiangsu, 211111, China (e-mail: jiaziye@nuaa.edu.cn).

Modern Large Language Models (LLMs) exhibit impressive zero-shot and few-shot generalization capabilities across complex natural language tasks, enabling their widespread use as virtual assistants for diverse applications such as translation and summarization. Despite being trained solely on large corpora of text without explicit supervision on author intent, LLMs appear to infer the underlying meaning of textual interactions. This raises a fundamental question: can LLMs model and reason about the intentions of others, i.e., do they possess a form of theory of mind? Understanding other's intentions is crucial for effective collaboration, which underpins human societal success and is essential for cooperative interactions among multiple agents, including humans and autonomous systems. In this work, we investigate the theory of mind in LLMs through the lens of cooperative multi-agent reinforcement learning (MARL), where agents learn to collaborate via repeated interactions, mirroring human social reasoning. Our approach aims to enhance artificial agent's ability to adapt and cooperate with both artificial and human partners. By leveraging LLM-based agents capable of natural language interaction, we move towards creating hybrid human-AI systems that can foster seamless collaboration, with broad implications for the future of human-artificial interaction.

Large language model based multi-agent systems have demonstrated significant potential in social simulation and complex task resolution domains. However, current frameworks face critical challenges in system architecture design, cross-domain generalizability, and performance guarantees, particularly as task complexity and number of agents increases. We introduces AgentGroupChat-V2, a novel framework addressing these challenges through three core innovations: (1) a divide-and-conquer fully parallel architecture that decomposes user queries into hierarchical task forest structures enabling dependency management and distributed concurrent processing. (2) an adaptive collaboration engine that dynamically selects heterogeneous LLM combinations and interaction modes based on task characteristics. (3) agent organization optimization strategies combining divide-and-conquer approaches for efficient problem decomposition. Extensive experiments demonstrate AgentGroupChat-V2's superior performance across diverse domains, achieving 91.50% accuracy on GSM8K (exceeding the best baseline by 5.6 percentage points), 30.4% accuracy on competition-level AIME (nearly doubling other methods), and 79.20% pass@1 on HumanEval. Performance advantages become increasingly pronounced with higher task difficulty, particularly on Level 5 MATH problems where improvements exceed 11 percentage points compared to state-of-the-art baselines. These results confirm that AgentGroupChat-V2 provides a comprehensive solution for building efficient, general-purpose LLM multi-agent systems with significant advantages in complex reasoning scenarios. Code is available at https://github.com/MikeGu721/AgentGroupChat-V2.