109 papers
Scale-Plan is a scalable framework that leverages large language models to filter irrelevant perceptual information and construct compact, task-relevant representations from natural language instructions, thereby enabling efficient and reliable long-horizon planning for heterogeneous multi-robot teams while outperforming existing baselines on the new MAT2-THOR benchmark.
This paper proposes a lightweight, training-free framework that parameterizes prompts as actions to dynamically influence LLM multi-agent dialogue behaviors, demonstrating through experiments that this policy-based approach effectively controls conversational dynamics across various indicators.
This paper proposes "Context Engineering" as a foundational discipline that, alongside Intent and Specification Engineering, forms a maturity model for scaling autonomous multi-agent systems by shifting focus from individual prompts to the comprehensive management of an agent's informational environment, goals, and policy constraints.
This paper models multi-LLM deliberation committees as random dynamical systems and demonstrates that even under deterministic temperature settings, factors like role differentiation and model heterogeneity induce chaotic instability characterized by positive empirical Lyapunov exponents, thereby establishing stability auditing as a critical requirement for reliable AI governance.
This paper introduces the LLM Delegate Protocol (LDP), an AI-native communication framework that enhances multi-agent system efficiency and governance by exposing model identity and reasoning profiles as first-class primitives, demonstrating significant reductions in latency and token usage alongside improved security and recovery capabilities in experimental evaluations.
This paper demonstrates that competing algorithmic agents using multi-armed bandit learning can spontaneously develop "naive collusion" through action synchronicity, even without knowledge of competitors, with the likelihood of such outcomes depending critically on whether the agents employ deterministic, greedy-in-the-limit, or persistently random policies.
This survey reviews the application of utility theory to cognitive modeling in robotics, tracing its evolution from behavior-based approaches to value systems that guide decision-making, learning, and cooperation in single and multi-agent environments, while identifying current limitations and proposing future research directions.
This paper demonstrates that explicitly reducing observation dimensionality and implementing locality-aware credit assignment in a communication-free multi-agent system enhances robustness and performance in asymmetric 3D pursuit-evasion tasks within cluttered environments.
Using an interdisciplinary multi-agent model, this study demonstrates that while climate change has a minor impact on Senegalese artisanal fisheries, reducing fishing effort is critical to preventing fishery collapse and mass migration, thereby enabling a sustainable equilibrium that restores historical catch levels.
TeamHOI is a decentralized framework that leverages a Transformer-based policy and a masked Adversarial Motion Prior strategy to enable a single unified policy to control scalable, physically realistic cooperative human-object interactions among any number of humanoid agents.
This paper addresses the challenge of agents with heterogeneous goals deciding when to cooperate or act alone by introducing a hierarchical learning framework that combines imitation and reinforcement learning, demonstrating superior performance over baselines and revealing that modeling teammates is most beneficial when their goals are less observable.
The paper introduces NarrativeLoom, a multi-persona collaborative storytelling system grounded in Campbell's theory of blind variation and selective retention, which a controlled study of 50 participants shows significantly enhances the novelty, diversity, and overall creativity of co-authored stories compared to existing tools, particularly benefiting novice writers through structured scaffolding.
This paper investigates concurrent graph games with distributed randomization where team players lack a shared random source, establishing that memoryless strategies suffice for the threshold problem (placing it in and proving NP-hardness) and that almost-sure reachability is NP-complete, while introducing the IRATL logic and a corresponding solver.
This study evaluates four multi-agent LLM topologies for rare disease diagnosis, revealing that while hierarchical structures marginally outperform single-agent baselines, increased complexity does not guarantee better reasoning and can significantly degrade performance through artificial doubt, thus supporting a shift toward dynamic topology selection.
This paper introduces MAS-Orchestra, a training-time framework that optimizes multi-agent system orchestration via function-calling reinforcement learning, alongside the MASBENCH benchmark, to demonstrate that multi-agent benefits are task-dependent and to achieve significant performance gains with over 10x efficiency on complex reasoning tasks.
The paper introduces FOR-Prompting, a model-agnostic, asymmetric prompting protocol that enhances reasoning and iterative refinement across diverse tasks by structuring interactions between a Defender, a Questioner, and an optional Host, enabling even small models to achieve performance comparable to or better than standard baselines without requiring training or access to model internals.
This paper establishes a comprehensive multi-KPI benchmark for Multi-Agent Reinforcement Learning in urban energy management using the CityLearn environment, demonstrating that Decentralized Training with Decentralized Execution (DTDE) consistently outperforms Centralized Training with Decentralized Execution (CTDE) in both average and worst-case performance while offering greater resilience and sustainability.
This paper introduces LatentMem, a learnable multi-agent memory framework that addresses memory homogenization and information overload by using an experience bank and a memory composer to generate customized, token-efficient latent memories, further optimized via Latent Memory Policy Optimization (LMPO) to significantly enhance multi-agent system performance.
The paper introduces SelfOrg, a training-free framework that enables multi-agent systems to self-organize by dynamically constructing response-conditioned communication graphs based on Shapley value approximations, thereby optimizing collaboration and significantly improving performance—especially with weaker LLMs—without relying on fixed topologies or external supervision.
Through large-scale experiments with over 1.5 million LLM-generated behavioral sequences, this paper reveals a fundamental asymmetry in behavioral inference where agent motivations are nearly perfectly recoverable while belief systems remain largely opaque due to inherent information-theoretic limits and architectural constraints, particularly within a "neutral zone" of behavioral ambiguity.