Multi-Agent Systems

As AI agents are increasingly deployed in complex environments, understanding their behaviors becomes critical. Yet behavioral scientific research on AI agents remains manual and labor-intensive. We introduce AEROBAT, the first multi-agent system to automate behavioral scientific research on AI agents. Given an arbitrary target behavior by its user, AEROBAT automatically executes a full pipeline of behavioral scientific research---generating hypotheses about the behavior, designing and executing controlled experiments, making behavioral assessments, analyzing the results, and writing reports. For 12 target behaviors, we used AEROBAT to generate and test 79 hypotheses: designing 1,240 controlled experiments and executing 23,512 simulation rounds in total. Moderate-to-strong statistical evidence was found for 26 hypotheses, including some novel ones. In sum, our results demonstrate that automated behavioral scientific research on AI agents can complement and extend the reach of manual research.

With advancements in long-distance wireless power transfer (WPT) and space-based energy technologies, integrating WPT into non-terrestrial networks (NTNs), referred to as NTN-WPT, is emerging as a promising approach for next-generation wireless networks. This paper proposes an energy-scheduling approach that jointly optimizes energy efficiency, task completion rate, and task waiting time for power transfer from low Earth orbit satellites to terrestrial mobile user devices (UDs). To address scheduling challenges caused by satellite and UD mobility and channel uncertainty from stochastic propagation effects, we decompose the problem into three subproblems within a three-layer predictive framework: 1) a state prediction layer forecasts UD and satellite states; 2) an interaction mapping layer uses a graph neural network (GNN) to model energy transfer efficiency; and 3) a decision-making layer determines the energy allocation plan. Distinct machine learning (ML) methods are tailored to each layer. To balance the competing objectives, we adopt a multi-objective reinforcement learning (MORL) technique that scalarizes them into a weighted-sum reward, transforming the multi-objective problem into a tractable single-objective problem. We further introduce a multi-agent deep learning model integrating self-attention with multi-agent proximal policy optimization (MAPPO) to improve objective balancing. Simulation results show that the proposed approach achieves a better overall trade-off than baseline methods, maintaining competitive task completion rates and energy efficiency while reducing task waiting times, and remains robust under highly variable conditions.

Multi-agent systems (MAS) built on Large Language Models (LLMs) are proliferating rapidly, but their heterogeneous execution traces provide no common basis for evaluation across methods. Outcome-only benchmarks discard collaborations, whereas LLM-as-Judge evaluation requires additional, model-dependent inference and can vary with the LLM and rubric. We introduce a generalizable evaluation framework that maps native MAS traces into a shared space of unified collaboration graphs, enabling different methods to be evaluated under the same representation, reference set, and metric panel. Candidate graphs are compared with a query-specific reference forest. Each forest is a benchmark-provided collection of verified-success graphs: it records diverse ways in which representative MAS methods can complete the task, rather than prescribing a unique optimal process. Instantiating the framework as ForestBench, we filter $844$ collaboration-necessary queries from seven public datasets, precompute ten successful target-conditioned reference graphs per query, and evaluate six representative MAS frameworks. Controlled backbone, reference-construction, and perturbation studies test the stability and scope of evaluation. Once the benchmark forests are built, ForestBench scores a trace in milliseconds without further LLM inference, providing a reusable structural basis for comparing diverse MAS collaboration traces.

Multi-agent reinforcement learning (MARL) is a powerful framework for solving complex collaborative tasks, but it relies heavily on well-defined global reward functions. Designing such rewards is challenging, especially in systems with heterogeneous agents, where a single scalar objective may fail to capture diverse behaviors. In this paper, we introduce Multi-AGent Preference-Integrated lEarning (MAGPIE), which addresses these challenges through agent-specific preference modeling. Each agent is evaluated by a dedicated expert through preference signals, eliminating the need for global evaluation. We theoretically prove that optimizing these decentralized preferences converges to a Nash equilibrium policy. To integrate local preferences into a coherent global objective, we construct agent-specific reward models from preference data and combine them via a monotonic aggregation mechanism. We further prove that optimizing this aggregate reward model is equivalent to training the Nash equilibrium policy. Extensive experiments on benchmark multi-agent tasks and a sequential production line task show that MAGPIE achieves performance comparable to reward-engineered baselines, demonstrating its potential to facilitate policy learning in scenarios where precise reward engineering is impractical.

An agentic organization should not be a set of model instances with corporate titles, despite most MAS still operationalizing organization as a conversational topology, a role prompt, or a fixed workflow. This paper develops an agent-native organizational structure framework that separates the persistent and dynamic layers of operations. The persistent layer consists of a four-store record architecture and a pool of resident specialization agents. A coordination layer defines Permission as the boundary of the operational world available to an agent, and Privilege as the set of organizational state changes that the agent is authorized to initiate. Together, these mechanisms compile task-specific operational worlds. A runtime layer combines an external Workflow Protocol with an isolated runtime store to dynamically assemble Task Groups. A human-interaction layer exposes the organization through a Control Plane mediated by a non-decision-making Translation Agent. Three orthogonal Role Groups further separate Operation, Review, and Supervision. Operators execute within narrowly scoped leases; reviewers receive elevated but demand-activated authority to modify organizational state; supervisors retain broad observational access while holding limited modification authority. The resulting architecture is fluid at the execution surface but structurally rigid underneath: tasks and events may alter team composition, topology, views, tools, and workflows, while records, write constraints, authority boundaries, and separation of powers remain persistent. The architecture has been implemented as a prototype and evaluated in small-sample experiments. Large-scale empirical validation remains incomplete, therefore no general performance claimed is made yet. Instead, the contribution is a coherent and falsifiable framework for designing, governing, recovering, and evaluating agent-native organizations.

Multi-agent LLM systems often fail not for lack of candidate answers, but because they have no persistent mechanism for deciding which incompatible claim should currently be trusted. Majority vote, debate, and judge-based selection choose an output without recording which claim wins, which is contested, or why a later update supersedes it. We present \term{LatticeMind}, a conflict-aware structured memory that handles contradiction at write time. It maintains explicit item status, applies cheap symbolic conflict checks, and invokes LLM reconciliation only for unresolved semantic cases. On a label-blind ConflictBank evaluation that removes source-name hints, LatticeMind reaches 0.97 accuracy versus 0.61 for the strongest aggregation baseline, with the gap significant at $p<10^{-6}$ by paired McNemar test. Ablations show that removing the checker or the reconciler costs 12 to 14 points. On four secondary planning benchmarks the picture is mixed: LatticeMind beats naive merge on three of four, but does not replace deliberation methods on tasks rewarding iterative search.

Robust perception under low-visibility conditions requires fused imagery that jointly preserves infrared thermal saliency and polarization-derived structural details. However, existing infrared-polarization image fusion (IPIF) methods often overemphasize dominant infrared responses, causing weak yet informative polarization textures in dark regions to be suppressed. To address this issue, we propose IRPol-Fuse, an energy-structure coordinated IPIF framework for challenging low-visibility scenarios. The proposed framework contains three key modules: Polarization Attention Fusion for adaptive infrared-polarization allocation, Infrared Highlight Injector for highlight-guided infrared preservation, and Polarization Texture Injector for polarization texture restoration and fine-detail recovery. We further construct LI-PI, a dedicated infrared-polarization evaluation dataset for low-visibility and visually concealed scenes. Experiments on LI-PI and the public LDDRS dataset demonstrate that IRPol-Fuse achieves favorable performance in thermal target preservation, structural detail recovery, and visual naturalness. Region-aware evaluation and downstream object detection further verify that the proposed energy-structure coordination strategy effectively preserves both infrared target saliency and polarization-derived structural information. Code is available at https://github.com/1hzf/IRPolar-Fuse .

Intermodal freight networks face growing disruption risk from climate extremes that degrade multiple corridors simultaneously. To address this, we formulate freight routing as a Dec-POMDP with per-shipment action granularity and train Independent PPO (IPPO) under Centralized Training with Decentralized Execution, comparing against two heuristic baselines with privileged state access on a 15-hub network under hurricane disruption. Across 30 matched episodes, no single policy dominates: IPPO achieves the highest throughput ($+12.7\%$) and delivery rate while a capacity-aware heuristic leads on Resilience Index (RI) and delay. Under demand surge (2.9:1 capacity ratio), IPPO's RI advantage grows to $+6.4\%$, suggesting learned routing is most valuable when capacity is scarce. A Multi-Agent PPO (MAPPO) variant collapses under train-eval queue mismatch ($\mathrm{RI} = 0.811$); retraining recovers RI to $1.018$ but IPPO still leads on throughput, pointing to residual limitations in centralized critics under per-shipment dispatch.

IT audits require auditors to judge whether heterogeneous organizational evidence satisfies semantic security and compliance controls. This judgment is difficult to automate because relevant evidence is distributed across policies, records, spreadsheets, and operational artifacts, and because audit conclusions depend on evidentiary sufficiency rather than keyword matching. We present IntelliAudit, a retrieval-grounded multi-agent system for IT audit evidence evaluation. Given a control and an evidence corpus, IntelliAudit retrieves relevant artifacts, generates an evidence-grounded assessment, challenges adverse findings, adjudicates disagreements, and produces an auditor-facing recommendation with cited evidence, rationale, missing-evidence analysis, and remediation guidance. We instantiate IntelliAudit on ISO/IEC 27001 and evaluate it across multiple simulated organizations using expert auditor review and audit-readiness user feedback. The evaluation shows that IntelliAudit can support control interpretation, evidence-grounded reasoning, and audit-preparation workflows, while also revealing the importance of human oversight for calibrating sufficiency judgments and correcting overly permissive recommendations. These results suggest that retrieval-grounded multi-agent systems can assist audit evidence review, but should remain decision-support tools rather than autonomous certification systems.

A common assumption when designing an agent in a multi-agent system is that the other agents behave adversarially. This allows a designer to obtain the strongest guarantees when they have no control over nor knowledge about the other agents' behavior. However, when all agents are designed under this adversarial assumption, their actual interaction is not adversarial (e.g., when all players play defensively, no player actually attacks). In such settings, we would like to know what behavior arises in the multi-agent system. However, analyzing the interaction among agents is notoriously challenging, both mathematically and algorithmically. In this paper, we provide such an analysis, focusing on bidding games, played by two agents on a graph as follows. A token is placed on a vertex, and in each turn an auction (bidding) determines which agent moves the token, thus generating an infinite path that determines the agents' utilities. We consider mean-payoff objectives; each vertex is associated with a reward for each player, and the utility in an infinite play is the limit average of the rewards. We analyze the play that is generated when each agent follows a strategy that optimizes against an adversary, and consider the two known explicit constructions of optimal strategies. The technical challenge stems from the infinitely-many configurations of a bidding game and their complicated dynamics. We show that, under some restrictions, the generated play is ultimately periodic, and develop algorithms to compute the players' utilities in it.

Musical interaction with AI is often organised as a response loop: a human performs, the system interprets that action, and the system answers, accompanies, or schedules a musical event. Unconducted vocal ensembles pose a different problem. Singers act simultaneously and continuously affect one another; neither timing nor pitch is fixed by a conductor, metronome, accompaniment, score, or tuning source. Collective organisation emerges from many-to-many reciprocal adjustment. This paper frames such ensembles as coupled dynamic systems and proposes a research architecture for vocal agents that enter, rather than merely track, their collective states. Some target repertoires are metrical, while others exhibit non-isochronous temporal contours that cannot be reduced to a beat grid; we treat the latter as a hard case for a general framework. The architecture connects multichannel capture in the field to dialect- and singing-aware representation, collective-state inference, vocal generation, and in-situ evaluation. The resulting agenda asks not only whether an artificial singer can synchronise, but how its presence reorganises human coordination, leadership, style, and musical transmission.

Multi-agent simulations are widely used to study complex social and ecological systems, where rich and often unexpected emergent behaviors arise from local interactions. A large body of prior work has focused on analyzing such emergent dynamics across domains. In this paper, we move beyond analyzing emergent behavior and introduce a learning-based mechanism for actively shaping it via social reward modeling. We introduce Multi-Agent Reward Prediction (MARP), a simple framework that extends preference-based reward modeling to multi-agent reinforcement learning. While the framework is designed to be applicable across multi-agent settings, the present empirical validation is limited to a single environment, and we therefore present MARP as a proof of concept within the studied domain. Rather than relying on handcrafted rewards, MARP learns a shared reward model from episode-level evaluations of collective outcomes, enabling decentralized agents to align their behavior with global social objectives. We study MARP in the Harvest Game, a canonical sequential social dilemma modeling common-pool resource management and related real-world challenges. Our results show that MARP can be tuned to produce behavior that is more closely aligned with target social metrics than standard reward-based baselines, while the learned reward model captures subtle environmental structure without explicit programming. Crucially, MARP supports multiple and composite social objectives within a single training regime. By modifying only the high-level evaluation metric, the same framework seamlessly aligns agent behavior with diverse goals, including sustainability, equality, and peace, as well as combinations of individual and group-level objectives. These findings demonstrate that emergent multi-agent behavior can be treated not only as a phenomenon to study, but as a target of principled, data-driven regulation.

Many methods for automated multi-agent system design optimize prompts and topologies during an initial design stage and then deploy the resulting system unchanged on subsequent samples. Experience from these samples is rarely consolidated into reusable system updates, while accuracy-oriented designs may incur high token costs. We introduce EMAS (Evolving Multi-Agent System), which uses this experience to revise MAS topology and prompts without updating LLM parameters, either to improve accuracy or to reduce cost. EMAS converts traces into structured diagnoses that specify a revision operation and target. It generates a candidate revision only when the same diagnosis recurs across samples and applies it only if paired validation against the current MAS meets the corresponding acceptance criterion. Across four benchmarks and two LLMs, EMAS attains the highest task-weighted overall accuracy for both backbones and is best or tied in six of eight model--benchmark settings. Within two evolution epochs, EMAS achieves relative gains of 6.30% and 20.10% in task-weighted accuracy on Kimi-K2-6 and Qwen3.6-27B, respectively. On MBPP with Qwen3.6-27B, EMAS raises accuracy from 55.09% to 89.12% while reducing token use per task by 62.2%. These results show that EMAS can turn experience from new samples into reusable updates to MAS topology and prompts.

Modern manufacturing imposes six coupled demands on adaptive control: local decisions with global consequences, partial observability, nonstationarity, reflex speed response with long horizon effects, delayed and diffuse outcomes, and dynamics that resist explicit modeling. Cooperative multiagent reinforcement learning (MARL), posed as a Dec-POMDP under centralized training with decentralized execution, is a particularly natural formalism for these demands. This paper adopts a MARL centered scope and asks where large language models (LLMs) should augment, interface with, train, or, in the strongest competitive case, replace that coordination core. A taxonomy organizes the literature through four LLM attachment points: policy, reward design, communication between agents, and hierarchical planning. A conditional capability profile separates native mechanism, reported performance, formal guarantee, and engineering maturity, and a deployment readiness analysis identifies the evidence behind each role. These stages yield the principal contribution: a three layer MARL centered reference architecture, grounded in evidence, for semantic reasoning, adaptive cooperative control, and independently assured execution. The LLM-Augmented Dec-POMDP is a descriptive comparative notation for that architecture, recording four attachment choices without introducing a new decision process class or algorithm. Under the reviewed evidence, conventional MARL is better suited to frequent, structured, decentralized coordination after task specific training, whereas LLM components are promising for semantic interpretation, reward drafting, human interaction, and slower supervisory planning. Current LLM only manufacturing controllers do not yet establish equivalence for strict real time, decentralized, safety critical control; this conclusion is bounded by the available evidence and does not assert impossibility.

Most CubeSats, small and low-cost satellites roughly the size of a shoebox, do not survive as long as they were designed to: a study of 178 missions found that only 48-65% remain operational after two years, against a designed lifetime of 2-5 years. The deeper issue is that a CubeSat in low Earth orbit (LEO) is physically unreachable from the ground for roughly 85 minutes out of every 96-minute orbit, so faults that start during that window go unnoticed until the next contact pass, by which point recovery may no longer be possible. We propose PHOENIX (Predictive Health On-orbit Edge Neural Intelligence eXtension) to give the satellite its own fault reasoning capability. A fine-tuned Small Language Model (SLM) compact enough to run on embedded hardware is deployed onboard the CubeSat, running on the flight-proven Aethero NxN-ECM computer, monitoring all sensor readings continuously, and resolving recurring faults using a memory system that stores past repairs so the same inference does not need to run twice. Once per orbit it sends a short structured health report to the ground instead of a raw data dump; six specialized AI agents on the ground read that report and generate validated satellite commands within the 5-10 minute contact window. A generative diffusion model (DDPM) creates synthetic training data because real fault examples make up only 0.57-1.80% of the dataset. We report preliminary results on the ESA Anomaly Detection Benchmark (14 years, 76 channels, 118 labeled faults).

Vision-Language-Action (VLA) models have recently achieved impressive performance for end-to-end autonomous driving, yet existing approaches are primarily designed for an individual single autonomous driving agent with limited support for cooperative perception, reasoning, and planning. We present Cooperative Multi-agent Unified Driving with Reasoning (CMU-Drive), a closed-loop end-to-end benchmark for evaluating cooperative autonomous driving with multiple connected autonomous vehicles (CAVs) operating in safety-critical driving scenarios with background traffic participants. We further propose Vehicle-to-Vehicle Vision-Language-Action (V2V-VLA), a cooperative VLA model that integrates cooperative driving into a single forward pass by jointly generating driving actions, future waypoints, language reasoning, and communication policies. Experiments on CMU-Drive establish the first benchmark and baseline for cooperative VLA driving and provide a foundation for future research on multi-agent, closed-loop, end-to-end cooperative autonomous driving. Our code, benchmark, and model checkpoint will be publicly released to facilitate open-source research.

The malicious use of generative artificial intelligence to create highly realistic deepfake videos raises serious ethical concerns and poses substantial challenges to AI safety. However, existing deepfake video benchmarks provide limited coverage of recent synthesis methods and generally lack reliable fine-grained textual annotations. Meanwhile, conventional detectors and multimodal large language models (MLLMs), whether operating as a single model or relying on a single analytical perspective, often fail to capture subtle forgery artifacts, limiting their generalization to emerging AI-generated methods. To address these limitations, we introduce FaceVid-Forensics-100K, a large-scale deepfake video dataset comprising 100,000 videos and spanning 33 synthesis methods across face swapping, face reenactment, and entire-face synthesis, including recent generators such as Seedance 2.0. The dataset provides fine-grained textual annotations of visual observations and verdict-consistent forensic explanations, automatically synthesized through a multi-model aggregation and conflict-resolution pipeline powered by advanced MLLMs. Building on this benchmark, we propose a multi-agent forensic reasoning framework that employs four specialized domain-expert agents to independently analyze forgery cues from four perspectives: texture, lighting, motion, and physics. A judge agent then reconciles their reports to produce a final prediction together with an explanation. Extensive evaluations on out-of-domain test sets show that, despite being composed entirely of small open-source MLLMs, our framework outperforms all methods including closed-source GPT and Gemini models and ranks first across all reported metrics on this benchmark. The project page is available at https://xavierjiezou.github.io/ARGUS/.

Large Language Models (LLMs) are increasingly used as high-level planners in embodied multi-robot systems, enabling robots to interpret natural language instructions and coordinate executable actions. Yet, this growing reliance on LLM planners also raises security concerns. Prior work has focused mainly on individual robots, while communication risks in multi-robot collaboration remain insufficiently understood. Existing multi-robot studies are further limited to preliminary analysis under the Decentralized Multi-agent System (DMAS) architecture, so it remains unclear whether these risks persist across other common communication architectures and how attacker access settings shape their propagation. To fill this gap, we formulate two communication attacks corresponding to distinct attacker access settings: the External Entry Point Attack and the Privileged In-System Attack. We evaluate both attacks across DMAS, HMAS-1, and HMAS-2 using three LLMs and five embodied multi-robot tasks. Results show that unsafe information can turn into unsafe actions across all three architectures: DMAS reaches a 96.7% entry endorsement rate and a 100% post endorsement activation rate, HMAS-1 reaches a 97.8% unsafe action success rate, and HMAS-2 triggers 88.3% of task defined unsafe action slots. To mitigate risks from trusted information flow, we introduce the Claim Provenance and Verification (CPV) Gate, which verifies communicated claims before downstream reuse and reduces the violation rate from 70.0% to 36.6%.

This paper presents a real-time Plan-and-Avoid (PAA framework for coordinating cooperative multi-agent airspace operations around a declared priority trajectory. The priority trajectory represents an aircraft flight plan that must be preserved because of constrained maneuverability, an emergency, a mission-critical task, or assigned operational priority. The framework predicts uncertainty-aware, well-clear separation violations with surrounding traffic and, when the priority plan alone cannot maintain separation, generates vehicle-constrained unilateral advisories that modify nearby aircraft trajectories to maintain well-clear separation for all traffic. The approach is applicable to any declared priority trajectory. This paper demonstrates the Plan component using a contingency landing planner to generate candidate priority trajectories. PAA then identifies nearby aircraft passing too close to this priority trajectory and issues Avoid resolution advisories to these aircraft. The framework is tested using real-world Automatic Dependent Surveillance-Broadcast (ADS-B) traffic from the Washington, D.C., airspace across more than 900 forced-landing cases, totaling over 140 hours of simulated flight. The PAA framework generates feasible cooperative advisories for all 575 unique conflict encounters, with a worst-case end-to-end response time of 5.7 s on a personal computer, including priority trajectory planning, advisory generation, and 1 s two-way datalink delay. In total, 93.5% of generated advisories satisfy the 35 s RTCA DO-365 Detect-and-Avoid temporal threshold. These results demonstrate low-latency coordination for preserving priority trajectories while maintaining well-clear separation through real-time automated advisory generation. Future work will quantify advisory-induced delays and their operational impacts.

We study online cooperative control of a multi-agent system under Byzantine attacks. Namely, an unknown, fixed subset of agents are Byzantine comprised and can stealthily overwrite its own coordinates of the team's planned joint action after observing that plan. The learner observes planned actions, public rewards, and public states, but neither the overwrite nor the executed joint action. Our objective is security: to optimize the team performance against the worst overwrites and achieve the optimal security value. We first show that the attacker's information determines the geometry. An attacker that observes the planned action induces an exact $(s,a)$-rectangular robust Markov decision process (MDP) whose rows are convex hulls of overwrite-induced public-outcome laws, whereas a blind attacker induces an $s$-rectangular model. We then identify the information-theoretic limit of security learning, showing that the security regret decomposes exactly into return regret against the response generating the data and a cumulative response gap $D_K$. Two indistinguishable horizon-one instances force $Ω(K)$ expected security regret while return regret is zero, showing that dependence on $D_K$ is unavoidable. Finally, we develop a stage-tied robust estimation-to-decisions learner and prove a regret bound of $\widetilde{\mathcal O}\!\left(H^2S\sqrt{AK}\right)+\mathbb E[D_K]$. Our studies thus provide comprehensive theoretical and algorithmic foundations of reliable multi-agent systems under Byzantine attacks.

Tool use transforms LLMs into agents that act beyond their training data, and for code-capable models, programmatic tool calling extends this further by replacing rigid JSON calls with scripts that chain and parallelize naturally. However, a systematic evaluation of tools as code on an established benchmark across current and prior model generations under real-world task conditions has not been conducted. In this work, we empirically compare programmatic tool calling (PTC) to native JSON tool calling across 14 language models on BFCL v4. In the programmatic tool calling paradigm, tools are exposed as typed Python stubs that the model invokes through code, with execution and results handled in a single agent turn. Programmatic tool calling matches or exceeds native JSON tool calling in 11 of 14 models on BFCL v4, with the GPT-5.6 family achieving a 10.6% improvement over the JSON tool calling baseline. Further, it matches or outperforms baseline in 13 of 14 models under parallel fan-out, and holds stable under context rot conditions where baseline degrades 2.3% on average. Our results demonstrate that programmatic tool calling is a viable and robust alternative to JSON tool calling, with performance tracking model capability across release generations.

Orchestrated collectives of large language model (LLM) agents that debate and vote are an emerging form of computational intelligence: the intelligent behaviour resides in the \emph{interaction}, not in any single agent. They improve task accuracy, yet remain black boxes at the system level: there is no principled test of convergence, no bound on the rounds needed, and no faithful account of what drove a decision. This paper develops a novel framework based on Koopman operator theory and validates its theoretical guarantees on multi-agent consensus dynamics. Treating the collective as one nonlinear dynamical system on a communication graph, we read its essential behaviour off the spectrum of its Koopman transfer operator, an exact linear representation of the nonlinear dynamics estimated from interaction traces. The spectrum yields three machine-checkable certificates: the sub-dominant eigenvalue $λ_2$ fixes the intrinsic timescale of reasoning and yields a convergence deadline computable \emph{before} the debate runs; its eigenvector names the coherent factions the collective reasons in, and $|λ_2|$ certifies when that explanation is valid; and the leading spectral coordinates form a compressed, auditable message basis. On an attention-consensus model, the deadline tracks observed convergence with log--log correlation $0.93$ and bounds it in 96% of 24 configurations; attribution is exact whenever the spectrum certifies metastability; eight of 32 coordinates preserve the decision at 99.7% fidelity; and a certificate learned from 15 debates held on 60/60 held-out debates. The study runs in minutes on a CPU, making spectral certification a practical layer for trustworthy collective reasoning.

Large language model (LLM)-driven multi-agent systems typically require multiple model invocations and complex coordination during inference, and their execution strategies directly affect system accuracy, latency, and computational cost. Parallel execution provides a means to improve inference-time efficiency. From the perspective of inference-time execution, this paper models parallelism in multi-agent systems as two distinct levels of decision processes: Replica Parallelism, which explores multiple complete solution paths at the task level, and Structural Parallelism, which enables concurrent execution within a single solution path through task decomposition. However, the roles of different forms of parallelism and their interrelationships still lack systematic study in terms of unified organization and coordination. We therefore propose TIPEX, a controllable execution framework that unifies these two levels of parallelism and coordinates their roles within the inference process under a unified execution semantics while supporting systematic combinations and analyses of different parallel strategies and parameter configurations. Systematic experiments on the GAIA benchmark demonstrate that inference-time parallelism can significantly improve accuracy and reduce end-to-end latency at the cost of increased token consumption. Further analysis shows that Replica and Structural Parallelism exhibit complementary effects across task complexities, with tasks of intermediate difficulty benefiting most from their coordination, while overly aggressive parallel strategies do not necessarily yield better performance.

Time-sensitive networking (TSN) is increasingly integrated into mobile edge computing (MEC) to support applications with stringent latency requirements, such as extended reality (XR). However, existing TSN scheduling solutions predominantly rely on static optimization techniques or centralized learning models that are based on fixed traffic patterns, limiting their effectiveness in dynamic environments. In practice, MEC environments often host multiple co-located XR traffic flows whose characteristics evolve over time, creating complex inter-queue dependencies that current schedulers fail to capture. Addressing these challenges requires adaptive, decentralized scheduling mechanisms capable of coordinating multiple TSN queues under varying traffic conditions. To this end, this paper proposes a multi-agent reinforcement learning (MARL) framework for TSN scheduling, where each TSN queue is modeled as an autonomous agent. The Heterogeneous-Agent Proximal Policy Optimization (HAPPO) algorithm is employed to explicitly model inter-agent dependencies and jointly optimize service delivery across queues. The simulation results demonstrate that the proposed approach reduces average frame waiting times by up to 26.8% and worst-case delays by approximately 16.8%, highlighting its effectiveness in dynamic XR-driven MEC scenarios.

Time-Sensitive Networking (TSN) and Mobile Edge Computing (MEC) hold strong potential for enabling ultra-reliable low-latency communication for time-sensitive applications, such as eXtended Reality (XR). However, the widespread adoption of XR introduces significant challenges due to co-located services in MEC environments, leading to contention for shared network resources. Moreover, XR traffic types have distinct characteristics and criticality in terms of timing requirements, further increasing the complexity and dynamics of such environments. Although reinforcement learning has shown promise for TSN scheduling optimization in dynamic network scenarios, existing approaches rely on centralized or high-level multi-agent designs and are typically tailored to periodic and predictable industrial traffic, limiting their applicability to XR workloads. As a result, these approaches suffer from (i) limited ability to capture inter-queue dependencies due to coarse-grained control, and (ii) poor adaptability to highly dynamic and heterogeneous XR traffic. To address these gaps, we propose a multi-agent reinforcement learning approach for queue-level XR traffic scheduling. We adopt the multi-agent transformer (MAT) to model inter-queue dependencies via attention over agents' observations and actions, enabling implicit coordination across heterogeneous co-located XR applications. Our simulation results show that the proposed method outperforms baselines, achieving up to 71.42% latency reduction and up to 83.2% reduction in failure rate, while consistently achieving high reliability across all queues.

Our results show that the existence of a short high-utility protocol already suffices for efficient communication. In particular, in a game with $n$ possible observations and $m$ actions: (1) For any achievable target utility $α$, we give an algorithm with $\mathrm{poly}(n, m, 1/ε)$ runtime that designs a protocol achieving utility at least $α-ε$ using only $2^{\mathcal O(CC_α(G))}/ε^2$ bits of communication. Here, $CC_α(G)$ is the minimum number of bits used by any protocol, even a computationally inefficient one, to achieve utility $α$. (2) We prove that this exponential dependence on $CC_α(G)$ is tight up to a constant. That is, unless $\mathrm P=\mathrm{NP}$, no polynomial-time algorithm can in general find optimal protocols using fewer than $2^{CC_α(G) -2}$ bits. We note that our results strictly weaken the assumptions required by prior work in the multi-agent information aggregation literature, filling a gap that had remained elusive even for games with constant $CC_α(G)$. In particular, prior guarantees for agreement-based information aggregation rely on structural assumptions such as informational substitutes or weak learnability. We show that these assumptions already imply $CC_α(G) = O(1)$ and are therefore more restrictive conditions than required by our protocol to succeed. On a technical level, our results involve a novel strengthening of the Frieze-Kannan weak regularity lemma and yield the following powerful polynomial-time transformation tool: for every communication game $G$, it constructs a game $\hat G$ that is a coarsening of the agents' observation spaces into constant-size partitions, such that $G$ and $\hat G$ are indistinguishable with respect to every short communication protocol. This coarsening theorem is the engine behind our algorithm and may be of independent interest.

We present PRIMAL3, an ultra-large-scale learning-based framework for multi-agent pathfinding (MAPF) that integrates reinforcement learning, topology-aware communication, LaCAM3-guided training, and PIBT-based action refinement. PRIMAL3 targets failures at topologically critical states, where agents must coordinate decisively around bottlenecks, dead ends, and persistent conflicts. Each agent is represented using features derived from cut vertices, dead-end regions, shortest-path distances, and blocking estimates. Two complementary graphs capture agent interactions: a same-direction following graph propagates multihop context along compatible paths, while a different-direction conflict graph differentiates agents competing for shared space through masked attention and relative features. During training, we propose to let policy entropy identify uncertain agents, for which LaCAM3 provides confidence-triggered action interventions and label-smoothed imitation targets. During execution, a priority-aware PIBT module refines the proposed joint actions using persistent, learned, and distance-aware priorities together with policy-aware fallback preferences while maintaining collision-free execution. The resulting framework combines learned exploration with structured expert guidance without requiring LaCAM3 at inference. Experiments demonstrate that PRIMAL3 substantially outperforms state-of-the-art learning-based baselines and scales to ultra-large instances with up to city-level 100,000 agents. Real-world experiments further demonstrate the feasibility of deploying PRIMAL3 on physical robotic systems and ablation studies validate the individual contributions the components we proposed. Project page: https://marmotlab.github.io/PRIMAL3/

Symbolic regression aims to discover closed-form equations from data, but existing LLM-guided methods often rely on a unified proposal loop that compresses heterogeneous search failures into a scalar score and a single prompt. We propose A-SR, a self-evolving agentic framework that shifts the control unit from expression edits to role-conditioned evidence views. A-SR coordinates formula discovery through routing among coordination protocols, an online evaluator-reward role policy, and state-routed process memory. During search, evaluator feedback characterizes reliability and productivity, updates role-level utilities, and routes elite motifs, failure traces, and validity diagnostics to different agents. The framework self-evolves at two timescales: within a run, it adapts the search process without updating LLM parameters; across runs, recorded trajectories can be distilled into open-source LLMs as role-conditioned proposal priors. Averaged over the four LSR-Synth scientific domains in LLM-SRBench, A-SR improves Acc@0.01 over baselines from 25.79% to 48.30% with Llama3.1-8B, while A-SR-LoRA improves the corresponding Qwen3-4B result from 24.58% to 38.29%. On four real-world scientific discovery tasks, A-SR obtains the best in-distribution or out-of-distribution normalized mean squared error on 7 of 8 reported metrics.

Cooperative multi-agent reinforcement learning often adds social terms to individual rewards, yet the scale of those terms is usually chosen by hand. We ask whether a guilt signal can instead be calibrated from human neural and behavioural data and transferred to artificial agents. Using the public SoDec responsibility fMRI dataset (40 participants), we fit a subject-fixed-effects regression of momentary-happiness changes on outcome-type counts and recover a guilt weight as the Partner-negative minus Social-negative contrast ($\hat{w}=1.118$, Cohen's $d=0.214$). We embed this weight in a two-agent Social Lottery environment and train independent Proximal Policy Optimization actor-critics under four shaping regimes: neurally calibrated, uniform constant, zero (selfish), and a unit-coefficient oracle. Across 1{,}000 evaluation episodes per condition, the calibrated agents track the human Social safe-choice rate most closely ($0.459$ vs.\ human $0.484$; $\mathrm{KL}=0.0012$), while the other three conditions deviate by one to three orders of magnitude in KL. Human neurobehavioural priors can therefore act as quantitative constraints on prosocial reward shaping.

LLM-based multi-agent systems (MAS) typically optimize a single topology, restricting reasoning to a narrow trajectory and limiting comprehensive analytical capacity. Naively merging multiple topologies into a composite graph introduces redundant noise propagation across irrelevant connections, degrading solution quality. To address this dilemma, we propose \textbf{Hierarchical Sparse Coordination over a Union of Complementary Topologies for MAS (HELENA)}, a multi-agent framework that balances diverse reasoning paths with sparse task-dependent execution. \helena{} constructs a union MAS graph from complementary candidate topologies selected via Monte Carlo Tree Search and Determinantal Point Process, broadening the reasoning trajectory for comprehensive analysis of complex problems. A Hierarchical Sparse Coordination module then activates only a sparse subgraph at each step while agents exchange compressed latent briefs to suppress redundant noise propagation. Finally, a Local Self-Refinement stage identifies decision units with discrepancy evidence and rewrites them only when contrastive evidence simultaneously confirms a reliable solution-side failure and a challenger-side improvement. Experiments across eight benchmarks show that \helena{} achieves state-of-the-art results on all benchmarks, with an average gain of \pctup{3.47} over the strongest baseline and up to \pctup{10.34} on MMLU-Pro, achieving larger improvements on harder benchmarks at a reasonable additional cost.

Can cooperation among large language model (LLM) agents be evolutionarily stable against free-rider invasion? We study an indirect reciprocity donation game where LLM agents observe behavioral traces and donate on a continuous scale. Strategies, represented as natural language prompts, evolve through cultural transmission across generations. Across four LLM backends, robustness to free-rider invasion varies by more than an order of magnitude. The strongest predictor of this robustness is opponent endowment sensitivity, the degree to which agents discriminate between cooperative and uncooperative opponents, operationalizing the classical Image Scoring mechanism. By contrast, adherence to the Leading-Eight L1 norm does not predict robustness. Robustness depends on defector exclusion: while both cooperator reward and defector punishment vary across models, only the stringency of defector exclusion predicts resistance to free-rider invasion. These findings reveal that LLM agents are confined to Image Scoring-like discrimination and fail to develop the more robust Leading-Eight norms, highlighting a fundamental vulnerability in culturally evolved LLM cooperation and motivating bottom-up approaches to norm construction.

Multi-agent LLM systems route among model-backed advisors, yet a deployer rarely knows before shipping whether routing will help at all. Prevailing routers optimize a gate's AUC and presume that advisor complementarity suffices. We show that neither determines the deployable gain. We introduce RouteGuard, a deployment-certification framework. Routing gain decomposes as $G = πΔ_E$, and the achievable gain is governed by a conditional-regret functional $Φ$, not by AUC. A finite-sample certification bracket comes with a matching Le Cam lower bound, constant-sharp over the fixed-activity class, and a robustness phase transition. On two benchmarks the framework acts as a guardrail. On RouterBench (11 cross-family models) the verdict depends on the sampling unit: the protocol certifies a gain over GPT-4 under prompt-level sampling and withholds it under workload-cluster resampling, because the gain rests on 3 of 86 workload cells. On OpenRCA (three Gemini advisors) the advisors are statistically redundant: the realized oracle sits at or below the independence baseline in all pools we tested (221 RouterBench pools and three OpenRCA distributions), so the protocol correctly refuses to certify. A pre-registered semi-synthetic control confirms calibration: the protocol certifies a genuine gain once $m \ge m^\star$ and does not certify a true null. Code and frozen artifacts will be released with the published version.

We present a structured large-language-model (LLM) architecture for zero-shot human--robot coordination in a cooperative construction task with private goal views. Guided by a Dec-POMDP formulation, the architecture decomposes decision-making into (i) action-conditioned Theory-of-Mind (ToM) inference, (ii) hierarchical planning, (iii) conversation interpretation, (iv) action verification, and (v) feedback-based replanning. We compare the proposed method with an ablation without ToM inference and a multi-agent reinforcement-learning policy trained offline over many goal pairs. In human-participant experiments, the proposed method required fewer interaction steps and yielded higher post-interaction trust ratings than both baselines. These results suggest that systematically decomposing the team decision problem, using LLMs as tractable surrogates for otherwise intractable inference and planning computations, and retaining conventional verification for physical feasibility can improve both task coordination and the human experience.

This paper proposes FedCritic-MIMO, a communication-efficient serverless federated multi-agent reinforcement learning framework for AI-native resource control across independently deployable cell-level controllers in open and disaggregated 6G RANs. Controllers share no trainer, retain local actors and personalized critic components, and exchange only compatible shared critic parameters. FedCritic-MIMO targets reuse-$1$ multi-cell massive-MIMO OFDMA deployments, where RAN controllers jointly manage user scheduling, per-stream power allocation, beamforming, interference, and long-term QoS with limited inter-controller signaling. Each base station locally executes its actor without centralized training or actor federation, while critic knowledge is exchanged peer-to-peer over an interference-aware graph. It enables this collaboration through wireless-aware event triggering, adaptive layer-wise top-$k$ sparse critic exchange with error feedback, and balanced interference-aware fusion. We establish conditional finite-time stationarity and consensus guarantees for the balanced, compressed peer-to-peer critic recursion under a fixed-policy, frozen-target critic-regression model. In strongly interference-coupled reuse-$1$ simulations, FedCritic-MIMO achieves the best performance-communication tradeoff among heuristic, independent-learning, centralized-training, and communication-ablation baselines. It achieves the highest held-out throughput, improves user-rate distribution and mean SINR, increases QoS satisfaction, and attains the lowest interference cost per delivered bit among learning baselines. It reduces critic-communication overhead by $76\%$ relative to uncompressed distributed critic exchange. These results demonstrate that serverless exchange of compatible shared critic parameters can coordinate RAN controllers without centralized trajectory collection or parameter-server aggregation.

AI agents are expected to play an increasingly important role in future decision-making systems. In this paper, we consider collaborative systems composed of heterogeneous multi-agent systems (MAS), where their members have different capabilities and operating costs. We study how agents can delegate tasks to one another so that certain research tasks can be completed effectively under resource-constrained scenarios. We first develop a multi-agent reinforcement learning-based (MARL) delegation training that enables agents to make sequential delegation decisions while minimizing the total execution cost. We then extend this approach to MARL with prescribed delegation topologies. Furthermore, we introduce two new frameworks for collaboration and delegation in multi-agent systems. The first framework proposes that an agent's policy depends not only on the current state of the underlying Markov decision process but also on the interaction history, including previous joint actions. This history-dependent formulation can improve coordination even in fully observable environments, where conventional MARL methods typically restrict policies to depend only on the current state. The second framework proposes a novel, potentially multi-dimensional monetary mechanism to facilitate the collaboration and delegation for MAS.

Clinical decision support is moving toward committees of language-model agents deliberating on a shared workspace. We ask whether such committees can be gamed by shortcuts, cues a benchmark rewards but a clinician would ignore. Across seven cohorts on six public datasets spanning text (MedQA-USMLE, MedMCQA, MIMIC-CXR reports), imaging (NIH ChestX-ray14, MIMIC-CXR-JPG, CheXpert) and tabular ICU records (SUPPORT2), Gemini committees resist these cues in isolation (flip 5-16%), yet a socially plausible shortcut spreads: when two peers assert the same wrong answer, the holdout under test adopts it in 38% of cases, as does a false "pre-screen" system flag, on both capability tiers. Of three oversight agents, a gate cannot separate adoption from honest agreement (false-positive rate 100%); a same-lineage judge reading only the transcript flags adoption on text (precision 100%, recall 93%) but collapses onto the gate in imaging; a referee that privately re-queries the holdout transfers to imaging (77-88% precision, 13-21% false-positive rate). Tripling a cue's visual salience does not move contagion, whereas a second peer voice raises it by half again. Gaming a hidden rubric is near-silent: only 1/10 text and 1/134 imaging drifters name the rubric they moved toward. What games a committee is social plausibility, and only a referee independent of self-report catches it. Code: https://github.com/criticaldata/benchmaxxing

AI agents deployed in real-world settings must be capable of coordinating with humans and other AI agents they have not encountered before. Zero-shot coordination (ZSC) algorithms aim to achieve this by specifying high-level learning rules such that independently engineered agents can coordinate with each other at test time. Rigorous evaluation of ZSC algorithms remains difficult: ideally, multiple independent implementations of each proposed algorithm must be used, reflecting the variation that arises when independent parties interpret and implement the same specification. In practice, however, ZSC algorithms have almost exclusively been evaluated using a single implementation trained across different random seeds, with only a handful of works additionally varying the neural network architecture. This leaves open questions about robustness to specification ambiguities and implementation details. In this work, we provide the first systematic evaluation of this robustness. We introduce a new evaluation scheme, cross-implementation cross-play, varying implementation details that prior work has shown to affect the performance of multi-agent reinforcement learning (MARL) algorithms, and we evaluate Other-Play, a popular ZSC algorithm, with this scheme. Our findings are encouraging and suggest that, for Other-Play, the standard ZSC evaluation is, in fact, a reasonable proxy for this more thorough cross-implementation evaluation.

Recent works train agents by constructing large-scale multimodal environment pools. However, we find that simply increasing the number of multimodal environments does not always benefit. We further analyze the limitations in current multimodal environment distributions through a series of experiments. Based on these findings, we study how to build more effective training environment distributions from two dimensions: diversity and difficulty structure. For diversity, we propose Ability-aware Environment Selection (AES) to obtain diverse environment sets. For difficulty structure, we propose Hierarchical Difficulty Curriculum (HDC), which organizes curriculum learning through two difficulty levels: harness weakening and state-scale progression. Experiments show that AES and HDC effectively improve multimodal agent training.

LLM-based multi-agent systems promise effective collaborative reasoning, but communication may amplify local errors into collective risks, and while existing evaluations emphasize final outcomes, they leave the reliability and propagation dynamics of distributed information aggregation unclear, so we introduce ForesightSafety-TIDE, a controlled evaluation framework that strictly pairs all-honest collaboration with controlled deception by a key evidence holder and analyzes the aggregation process through multi-stage voting, testimony adoption, and evidence-root lineage propagation, and using 120 five-agent object-movement environments where partial observations jointly determine a unique endpoint, we evaluate 3 homogeneous LLM-based multi-agent systems, and across these paired conditions, aggregate truth recovery falls from 72.50% to 14.17%, with significant declines for every system, while process tracing and exit ablations show that a single false testimony is adopted more readily than truthful testimony, propagates to higher orders, and persists through honest agents after the deceiver exits, and observers without first-hand evidence suppress incorrect consensus but do not improve truth recovery, so together, these findings reveal both the fragility of distributed fact recovery and its underlying mechanism: false evidence gains collective influence through its adoption and continued propagation by other agents after entering communication.

Enterprise forecasting increasingly relies on autonomous agents that interpret documents, search for data, generate code, and revise models. While this autonomy helps build adaptive forecasting pipelines, it also makes it difficult for practitioners to inspect why a forecast changed, which evidence supported the change, and how data and modeling choices were revised. We present TraceMAS, an interactive demo system for traceable multi-agent forecasting. TraceMAS organizes agent outputs around two causal-loop representations: an Ideal Causal Loop Diagram (Ideal CLD), which captures key factors and their causal relations extracted from domain documents, and a Data-Grounded Causal Loop Diagram (Data-Grounded CLD), which links those factors to internal variables, external data, or documented proxies. The Data-Grounded CLD guides feature construction and model design while preserving the connection between textual evidence, data choices, and model revisions. We demonstrate TraceMAS on crude oil price forecasting. The demo interface allows users to compare forecasting iterations, inspect agent-level revisions, explore causal maps, review feature-data mappings and model architecture, and connect scenario forecasts to market narratives. This demonstration shows how autonomous forecasting agents can retain flexibility while making the evidence-to-forecast process inspectable.

Large language model-based multi-agent systems (LLM-MAS) are designed through roles, debate protocols, and aggregation rules. These choices create implicit social expectations: agents may be expected to trust, challenge, defer to, or collaborate with peers. We study the effects of making inter-agent relation semantics explicit. We use a minimal signed-network formulation of relational priors and inject natural-language renderings into agent system prompts while holding the task protocol fixed. Across a commons-governance simulation and multi-agent debate, relational priors primarily act as convergence pressure: increasing relational positivity tends to make agents coordinate or agree more readily. This pressure can help when utility rewards behavioral alignment, as in sustainable resource governance and subjective consensus. It does not, however, reliably improve accuracy. In objective QA debates, higher positivity can increase agreement even when correctness-conditioned agreement does not improve and may decline in some settings. Effects vary by model backbone, relation type, and topology; explicit neutrality is not equivalent to omitting relational framing. We argue that relational priors should not be a default add-on for LLM-MAS. Their safer use is diagnostic and task-specific: compare against a no-prior baseline, monitor correctness-conditioned metrics when truth matters, and omit the relational layer when validation does not justify it.

Revenue forecasts guide acquisition budgets, demand planning, and customer-based valuations, yet an aggregate forecast does not show whether change reflects acquisition, repeat purchasing, spending per order, or offsetting movements. Using weekly transaction panels for 966 companies in 25 industries, the authors develop the Customer-Based Multi-task Transformer (CBMT), which learns shared structure, retains separate primitive forecasts, and aligns their combination with downstream revenue. CBMT's mean total-sales error is 30% below the strongest representative established customer-base benchmark. It is also 2.65% below a Transformer that forecasts total sales directly, although the paired difference is not statistically significant (p=.222), and it beats separately estimated single-task forecasts for 74.3% of firms. CBMT's source MAE is lower in 23 of 24 benchmark-by-outcome comparisons, with the remaining difference not statistically distinguishable from zero. Firms whose primitives co-move more strongly are more likely to benefit from joint forecasting; selected-family scenario-3 comparisons are consistent with gains from shared representation and revenue alignment but remain diagnostic rather than causal. Accuracy deteriorates for all models when customer-base dynamics are highly volatile, and CBMT's advantage narrows there. Calibration-period routing rules do not improve average accuracy over always deploying CBMT. The results show how coordinated customer-base forecasts support revenue planning and when they warrant greater caution.

Large language models have shown promise for automated Verilog RTL generation, yet state-of-the-art multi-agent systems plateau at ~95% accuracy on standard benchmarks. We trace this ceiling to an incomplete debugging action space: existing systems restrict which signals the agent can inspect, which time windows it can query, or both, reducing debugging to pattern matching on a narrow, predetermined view of circuit behavior rather than hypothesis-driven root-cause analysis. We present VeriTrace, a multi-agent system whose Inspector agent operates over a complete debugging action space, with independent control over signal selection, time-window bounds, and iteration depth. This capability, which we term Agentic Temporal Exploration, enables the agent to form hypotheses about failure causes, query the waveform for evidence, and refine its understanding iteratively, mirroring the exploratory process of human verification engineers. VeriTrace achieves 100% Pass@1 on VerilogEval-V2, the first system to attain perfect functional correctness on this benchmark. On a shared Claude Sonnet 4.0 backbone, VeriTrace outperforms the strongest reproduced baseline by +5.1%, demonstrating that debugging agency closes the final accuracy gap.

Multi-agent reasoning (MAR) improves reasoning reliability through iterative solution exchange and refinement. Existing adaptive MAR methods typically learn routing decisions from query-level labels or trajectory-level returns, but such coarse supervision cannot accurately estimate the state-conditioned utility of individual operators in multi-step collaboration. We propose TreeCredit, a shared-prefix credit assignment framework for efficient adaptive MAR. Its core insight is to estimate operator utility through state-matched downstream comparisons, rather than directly attributing trajectory-level outcomes to preceding decisions. TreeCredit constructs shared-prefix collaboration trees by expanding candidate operators from the same intermediate state and assigns each state--operator pair a correctness-prioritized suffix credit based on the terminal correctness and cumulative additional cost of its complete continuation. These structured credits are converted into state-local operator preferences to train a lightweight pairwise state router, which dynamically selects the next admissible operator during inference. Experiments on six reasoning benchmarks show that TreeCredit modestly improves accuracy while substantially reducing inference cost, achieving a better accuracy--cost trade-off than representative MAR methods.

Tool-using agents built on large language models (LLMs) are increasingly deployed not by a single operator but by many, side by side on shared infrastructure. This creates a population-level risk that single-agent safeguards miss: a handful of agents can quietly coordinate, rigging a market, boosting one another in a review process, or timing a joint data grab, while each one looks perfectly well-behaved. The difficulty is that the organisations running these agents cannot see inside one another's models, so any realistic detector must work from behaviour alone: black-box, trace-only, and often with only partial visibility. We treat covert coordination as an information-hiding problem and build a black-box steganalysis detector that combines cross-run mutual-information estimation, permutation tests, distributional-shift statistics, and timing and tool-call side channels, all calibrated to a fixed false-positive budget. Our central move is to stop testing against a single fixed code: we pit the detector against an adversary that continually rewrites its encoding to slip past whatever the detector has learned, and we run this red-versus-blue contest in tool-using, memory-carrying environments rather than toy games. Capacity theory then tells us what to expect, a detection-capacity frontier, a covert bit-rate below which black-box detection is provably no better than chance. We set out an experiment to map this frontier, report clearly labelled placeholder results pending measurement, and flag a practical evasion, spreading a payload across sessions, that current methods largely miss.

LLM multi-agent systems suffer from Correctness Collapse, where high task-level accuracy conceals severe process-level failures. This is especially hazardous in safety-critical embodied settings such as automotive cockpits, where lexically correct utterances may trigger dangerous physical operations. Existing attribution methods rely on text traces alone, missing dependency structure, multi-channel evidence, and safety-aware evaluation. We introduce CockpitHAT, a hierarchical attribution framework that replaces positional windows with dependency-distance thresholds from interaction DAGs, integrates multi-channel evidence via an embodied adapter, and applies a safety-uplift to high-risk failures during confidence-weighted analyst consensus. We further release CockpitBench, a benchmark of 212 annotated failure traces spanning dialogue, vehicle-state, environmental, and memory channels, each labeled with ISO 26262 ASIL severity via three-expert consensus. On the public Who&When benchmark, CockpitHAT achieves agent-level / step-exact accuracies of 77.9% / 37.8% on the Hand-Crafted split and 86.5% / 46.0% on the Algorithm-Generated split, surpassing the text-only SOTA ECHO by up to 17.6 / 16.7 points. On CockpitBench, it attains 78.3% agent-level and 38.2% step-exact accuracy. These results establish dependency-aware, multi-channel, risk-calibrated attribution as an effective paradigm for reliable failure diagnosis in real-world embodied LLM multi-agent systems.

Maximizing throughput under proportional fairness in dense wireless networks requires jointly managing user association, scheduling, base station (BS) activation, and handover control under hard finite-horizon energy and handover budgets, which induces a fundamental tension between BS-side energy management and user-side handover regulation. While multi-agent reinforcement learning (MARL) is a natural framework for such distributed sequential control, its application here faces two difficulties: finite-horizon budget constraints cannot be evaluated at each time slot, and the nonlinear proportional fairness utility admits no principled per-slot decomposition. We propose HeLyMARL, a Lyapunov-embedded heterogeneous MARL framework that resolves both via drift-plus-penalty decomposition with virtual queues. The energy and handover constraint pressures are internalized directly into a unified per-slot reward, converting the constrained finite-horizon problem into an unconstrained MARL problem. Comparison against two Lagrangian-based alternatives reveals a timescale separation: Lagrangian relaxation regulates constraints only across training episodes, whereas the virtual queues of HeLyMARL bound cumulative budget consumption at every partial horizon within an episode, a pacing guarantee beyond the reach of greedy Lyapunov-based control. Simulations show that HeLyMARL is the only method that sustains the throughput-fairness balance together with uninterrupted service throughout the horizon, outperforming conventional MARL, Lyapunov-based, and constrained MARL benchmarks without premature budget exhaustion.

LLM-based multi-agent coordination faces a fundamental trade-off between efficiency and adaptivity in dynamic environments. Existing approaches typically rely on repeated LLM invocations or multi-round communication to adapt decisions during execution, introducing substantial latency and making coordination vulnerable to asynchronous progress and environmental changes. Conversely, one-shot planning reduces coordination overhead but produces open-loop plans that can quickly become stale or fail when actions depend on other agents and the environment. We introduce SyncPlan, a plan-execute-correct framework for long-horizon coordination through explicit synchronization and adaptive correction. Given the state and team-level task, a centralized LLM coordinator generates per-agent action chains in a single planning call. During execution, explicit wait primitives and deadlock detection enforce inter-agent and agent-environment dependencies, while a lightweight Plan Staleness Detector continuously assesses the remaining plan and triggers replanning when environmental changes invalidate its assumptions. We further optimize the coordinator through SFT and planning-oriented RL with dense task progress and outcome-level execution feedback. Experiments on the public Overcooked benchmark and the complex Honor of Kings environment show that SyncPlan achieves state-of-the-art task success rates while using less than 0.05% of the wall-clock runtime compared with existing LLM-based coordinators. Code and datasets will be made publicly available.

This paper studies feedback Nash equilibrium (FBNE) seeking for multi-agent trajectory planning in nonlinear dynamical systems with unknown agents' objectives and state-dependent inter-agent coupling. While dynamic game theory provides a principled framework for such problems, existing approaches typically assume fully rational agents with known objectives or rely on fixed regularization, limiting their ability to capture bounded rationality and spatially varying interaction intensity in safety-critical settings. To this end, we propose a KL-regularized dynamic game with a state-dependent weight that adaptively balances optimality and behavioral priors. To infer unknown cost parameters from demonstrated behaviors, we develop a context-aware inverse game module based on maximum-entropy inverse reinforcement learning with physics-informed regularization, ensuring structural consistency with the forward game. We establish per-iteration well-posedness of the regularized local game and show that the adaptive weighting function remains Lipschitz continuous under bounded nominal-trajectory updates. Numerical simulations and multi-robot experiments on cooperative navigation and merging scenarios validate the effectiveness of the proposed framework.

Autonomous agents are increasingly used to execute consequential tasks in environments governed by operational constraints, organizational policies, regulatory requirements, and technical standards. Their safety is therefore determined not by the correctness of individual actions, but by whether their overall behavior remains consistent with the rules and invariants of the systems in which they operate. As large language model (LLM)-based agents become more autonomous and increasingly delegate tasks across organizational boundaries, securing them evolves from a single challenge into a broad and interconnected landscape spanning the entire agentic stack. At the single-agent level, untrusted inputs through prompts, memory, retrieved knowledge, and tool interfaces create attack surfaces. In multi-agent settings, delegation and communication introduce challenges related to identity, trust, capability control, and decision transparency, while the underlying model routing and execution control plane remains vulnerable to manipulation and to unverified model provenance. Perhaps the most fundamental challenge is behavioral containment: sequences of individually permissible actions may collectively violate system-level constraints and safety invariants. At the broader level, supply-chain integrity, provenance, accountability, and end-to-end observability remain largely open problems. A common principle unifies these directions: security must become a verifiable property of the architectures, protocols, and runtimes that govern agent behavior, rather than an optional layer of guidance. Charting these challenges provides a roadmap toward trustworthy autonomous agent deployment.

Training LLM-based multi-agent systems with multi-agent reinforcement learning is rapidly gaining traction, and a parallel line of work argues that such systems should be judged by their behavior, not only their reward. We take up both threads in spatial cooperative games, where small frozen LLMs prompted with low-level actions fail outright, earning zero reward. Guided by the options/semi-MDP framework---and, because option execution is asynchronous across agents, its multi-agent extension in macro-action Dec-POMDPs---we equip each game with a library of symbolic \emph{options}: typed, state-feasible, short-horizon behaviors executed by a symbolic planner. Each library is drafted by a frontier coding model from the game's source code; the feasibility guards that filter each menu are then synthesized mechanically from cheap random-policy burn-in rollouts---a guard is adopted only if it explains repeated execution failures while hiding no logged success---so no guard is authored, selected, or reward-tuned by hand. Each agent's LLM acts as its policy over options, with a private per-agent LoRA adapter trained by a per-agent variant of multi-agent GRPO (PA-MAGRPO); this lifts frozen bases from zero reward to competent play across three games and four small backbones. Behavioral audits then reveal that reward and cooperation decouple: a rising reward curve may simply mean that one agent has learned to run the entire task alone while its partner idles---cooperation emerges only when the task makes it necessary. Reward alone is thus an unreliable readout of cooperation; behavioral evaluation must sit alongside it.

An AI development race creates a multi-agent safety dilemma. Each company can develop slowly and safely, or move faster while taking a risk that may remove its final reward. We use this repeated game to study strategic safety behaviour among large language model (LLM) agents in races with two to five players. However, a valid action does not show that an agent understands the game. We therefore place an audit gate before behavioural interpretation. We first verify the game engine, then test rule recall, state tracking, payoff calculation, and stability under different but equivalent task descriptions. We then compare LLM action sequences with an evolutionary game-theory benchmark and published human data, and explore differences across models, risk conditions, personas, and two- to five-player races. The audit shows that strong rule recall can coexist with weak state tracking and expected-payoff calculation. Providing verified arithmetic and changing the response representation can also change later actions, even when the game rules stay fixed. Across seven tested model endpoints, aggregate rates hide large differences in action sequences, responses to opponents, and responses to race position. Patterns across the tested three- to five-player races are also model-specific rather than a single effect of adding competitors. These results show why multi-agent AI-race simulations need validity checks and trajectory-level analysis before their outputs are described as strategic, human-like, or safety-aware. Our findings are exploratory and apply only to the tested models, prompts, and decoding settings.

Evaluating Multi-Agent Reinforcement Learning (MARL) policies in autonomous driving fundamentally relies on extrinsic statistical indicators (e.g., reward curves and success rates), which often mask intrinsic policy degradation and algorithmic blind spots. To break this black-box evaluation, this letter proposes a novel information-theoretic diagnostic framework. By leveraging a fully converged Monte Carlo Tree Search (MCTS) as an asymptotic oracle, we establish a theoretical ground-truth baseline distribution. We formulate a bounded policy optimality score ($\mathcal{M}_{opt}$) using the forward KL divergence to rigorously penalize fatal collaborative omissions. Crucially, we semantically decouple this metric into lateral and longitudinal dimensions, creating a granular "semantic microscope". Extensive spatial and temporal diagnostics on state-of-the-art MARL architectures and exploration mechanisms demonstrate that our framework conclusively exposes hidden directional biases, identifies temporal average-policy traps, and transforms heuristic hyperparameter tuning into a visually trackable trajectory optimization. This framework establishes a rigorous, model-agnostic standard for benchmarking intrinsic multi-agent policy quality.

In post-disaster scenarios, unmanned aerial vehicles (UAVs) are critical for establishing emergency communication networks. For time-critical rescue missions, information freshness is crucial because decisions based on outdated data may lead to ineffective control actions. This paper investigates age of information (AoI) minimization for UAV-assisted emergency communications with heterogeneous emergency services. We model bursty packet arrivals using a Markov-modulated Poisson process and adopt finite blocklength theory to capture the coupling among transmission duration, packet completion, and AoI evolution. To balance delay-tolerant long-packet transmission and urgent short-packet response, we propose a mini-slot-embedded scheduling mechanism with adaptive checkpoint-interval selection. We formulate the joint optimization of UAV trajectory control, user scheduling, and checkpoint-interval selection as a multi-agent decision problem, and develop MA-HEAD-Net, an adaptive rule-guided multi-agent deep reinforcement learning framework. MA-HEAD-Net incorporates communication-domain rule priors into a gated multi-head policy, where adaptive gates regulate the contributions of rule-prior and learned-policy logits for different subtasks. The policy and gating components are jointly optimized under multi-agent proximal policy optimization. Simulation results show that MA-HEAD-Net improves policy-formation efficiency compared with representative multi-agent deep reinforcement learning baselines and achieves lower AoI than both learning-based and heuristic methods in dynamic UAV-assisted emergency communication scenarios.

LLM-based multi-agent systems (MAS) extend LLM capabilities through iterative communication and shared contexts. However, this collaboration introduces a vulnerability: backdoor behavior can be activated when peer evidence reaches a hidden threshold, rather than being determined by any single message. We introduce a collective evidence-threshold backdoor paradigm for MAS and Boundary-Conditioned Backdoor Injection (BCBI), which constructs counterfactual boundary pairs to separate benign behavior before the threshold from the adversarial objective after it, and learns latent progression aligned with evidence. To mitigate this threat, we propose LAtent Transition Test-time Evaluation (LATTE), a clean-only latent-transition defense that learns benign communication dynamics and quarantines anomalous agent updates before their responses propagate. Across several benchmarks, BCBI yields selective activation with little premature activation; without knowing the attack target or trigger, LATTE limits propagation with minimal disruption.

The emergence of Agentic AI systems, characterized by autonomous reasoning, multi-agent collaboration, tool orchestration, adaptive decision-making, and persistent memory, represents a fundamental shift from traditional AI pipelines to dynamic software ecosystems. While AI Technical Debt (AITD) has been widely studied in machine learning and software engineering, existing models assume static, component-level architectures and fail to capture the dynamic and emergent behaviors of agentic environments. To address this gap, this paper introduces Agentic Technical Debt (AgTD), defined as technical debt that emerges, accumulates, propagates, and amplifies due to the autonomous and collaborative nature of Agentic AI systems. Building on our prior systematic scoping review of 31 AITDs across seven root-cause categories, we employ a theory-informed transformation methodology to reinterpret these debts in Agentic AI through direct transformation, contextual transformation, and manifestation expansion. We present the first systematic mapping of established AITDs to their agentic manifestations, showing how conventional debts evolve into system-level liabilities, including memory inconsistencies, orchestration fragility, cascading failures, and unsafe autonomous decision-making. Our findings show that technical debt extends beyond software artifacts to encompass agent behaviors, coordination mechanisms, and interactions among agents, tools, and execution environments. We further examine its implications for AI Trust, Risk, and Security Management (AI TRiSM), highlighting impacts on trustworthiness, governance, security, operational resilience, and Sustainability Technical Debt. Overall, this work establishes AgTD as a foundational software engineering construct and provides a transformation framework, taxonomy, and research agenda for managing technical debt in autonomous multi-agent AI systems.

Multi-agent systems (MAS) decompose long-horizon tasks across supervisors and subagents, but delegated goals do not necessarily carry their original authorization boundaries. Existing safety benchmarks mainly study adversarial compromise, while work on constraint drift lacks controlled architecture-level evaluation. We introduce MasDrift, a benchmark of 600 benign productivity tasks across eight domains. Each task pairs required work with reserved actions. MasDrift compares single-agent, centralized, and decentralized coordination while varying hierarchy depth and peer width, measuring task completion and authorization preservation. Across generic multi-agent conditions, centralized hierarchies achieve 93.9--98.6% task completion versus 85.7--87.0% for peer networks, while unauthorized actions occur in 2.7--19.8% of tasks versus 0.6--0.8%, a gap that widens with hierarchy depth. We further compare two defenses that differ in where authorization evidence resides. One re-anchors every pending call to the original user request. The other carries an attenuated policy along the delegation chain. Re-anchoring reduces unauthorized actions in every model configuration we evaluate, at a cost of 1.6 points of pooled completion. Chain propagation blocks required work instead, forfeiting up to 36.3 points. A heterogeneous case study confirms that the failure follows from coordination rather than model strength. MasDrift exposes a centralization tradeoff and makes authorization preservation a measurable property of MAS design.

Autonomous AI agents, increasingly empowered by large language models, are becoming important components of human-machine systems for high-stakes decision support in digital twin ecosystems. However, existing multi-agent systems often lack robust verification for identity, capability, and policy compliance, especially in decentralized environments spanning multiple institutions. This paper proposes a neuro-symbolic decentralized governance framework for verifiable agents in collaborative digital twin environments. By representing agents through multi-layer semantic profiles, the framework bridges probabilistic neural reasoning with deterministic institutional governance, thereby supporting trustworthy human-AI collaboration and meaningful human oversight. Capabilities are grounded in formal domain ontologies to enable machine-interpretable, policy-aware, and context-sensitive participation. These credentials, issued by organizational authorities, are validated via blockchain-based smart contracts, ensuring auditable participation without exposing sensitive data. We demonstrate the framework using a decision-support prototype with clinic, digital twin, and wearable provider agents effectively prevents unauthorized interaction and enforces institutional policies with manageable overhead. Our findings suggest that neuro-symbolic decentralized governance provides a scalable and trustworthy pathway for safe human-machine collaboration across institutional boundaries.

Large language models are increasingly integrated into autonomous robotic systems for task planning and control, but this integration exposes them to prompt injection attacks that can lead to unsafe decisions and physical harm. Multi-agent settings increase the risks through cross-agent contamination and broader attack surfaces. In this paper, we evaluate prompt injection attacks against an LLM-based multi-agent robotic system, considering both direct injections into task instructions and indirect injections through perception modules. In our experiments across varying attack-goal complexities and injection strategies in both single-agent and multi-agent settings, we show that prompt injection can induce adversarial actions while reducing task completion. We find that attacks can propagate from one agent to others through shared prompt structures, with impacts varying depending on prompt composition and the targeted agent. We further analyze how architectural changes affect LLM queries and, consequently, the attack success. To the best of our knowledge, this is the first study that systematically investigates prompt injection attacks in a multi-agent LLM-based robotic system.

Safety interventions for large populations of network-coupled agents must protect shared constraints without unnecessarily overriding task-oriented policy decisions. We present HetGPS, a hybrid graph-control framework synergizing learned graph risk with physics-anchored correction by separating intervention magnitude from corrective direction. An action-conditioned graph residual model schedules state-dependent intervention authority, while a physics model determines its direction. For electric vehicle (EV) charging, we couple this filter with a parameter-shared heterogeneous graph soft actor-critic policy, enabling topology-aware coordination with a learned model size independent of fleet size. Across five nested distribution networks with 200--3,218 EVs and 100 evaluation days, Adaptive Authority reduces bus--step voltage violations from 3.93--7.74% without filtering to 0.52--3.44%, while maintaining 99.06--100% departure success. Relative to the same physics-directed projection with fixed authority, it improves mean reward on all five networks and lowers the mean safety score on four. The deployed policy-and-risk model contains 383,702 learned parameters at every scale; at 3,218 EVs, a matched centralized SAC actor is about $170\times$ larger. A policy trained on the eight-transformer system transfers zero-shot to the 16- and 32-transformer systems, attaining 0.57--0.75% violation rates and at least 99.99% departure success. These results show that learned graph risk can allocate intervention authority at scale while feeder physics anchors corrective action.