Independent on-policy policy gradient algorithms are widely used for multi-agent reinforcement learning (MARL) in cooperative and no-conflict games, but they are known to converge suboptimally when each agent's policy gradient points toward a suboptimal equilibrium. In this work, we identify a subtler failure mode that arises \textit{even when the expected policy gradients of all agents point toward an optimal solution.} After collecting a finite set of trajectories, stochasticity in independent action sampling can cause the joint data distribution to deviate from the expected joint on-policy distribution. This \textit{sampling error} w.r.t. the joint on-policy distribution produces inaccurate gradient estimates that can lead agents to converge suboptimally. In this paper, we investigate if joint sampling error can be reduced through coordinated action selection and whether doing so improves the reliability of policy gradient learning in MARL. Toward this end, we introduce an adaptive action sampling approach to reduce joint sampling error. Our method, Multi-Agent Proximal Robust On-Policy Sampling (MA-PROPS), uses a centralized behavior policy that we continually adapt to place larger probability on joint actions that are currently under-sampled w.r.t. the current joint policy. We empirically evaluate MA-PROPS in a diverse range of multi-agent games and demonstrate that (1) MA-PROPS reduces joint sampling error more efficiently than standard on-policy sampling and (2) improves the reliability of independent policy gradient algorithms, increasing the fraction of training runs that converge to an optimal joint policy.

In question-answering (QA) systems, Retrieval-Augmented Generation (RAG) has become pivotal in enhancing response accuracy and reducing hallucination issues. The architecture of RAG systems varies significantly, encompassing single-round RAG, iterative RAG, and reasoning RAG, each tailored to address different types of queries. Due to the varying complexity of real-world queries, a fixed RAG pipeline often struggles to balance performance and cost efficiency across different queries. To address this challenge, we propose an adaptive RAG framework called MAO-ARAG, which leverages multi-agent orchestration. Our adaptive RAG is conceived as a multi-turn framework. Specifically, we define multiple executor agents, representing typical RAG modules such as query reformulation agents, document selection agent, and generation agents. A planner agent intelligently selects and integrates the appropriate agents from these executors into a suitable workflow tailored for each query, striving for high-quality answers while maintaining reasonable costs. During each turn, the planner agent is trained using reinforcement learning, guided by an outcome-based reward (F1 score) and a cost-based penalty, continuously improving answer quality while keeping costs within a reasonable range.

The proposal introduces an innovative drone swarm perception system that aims to solve problems related to computational limitations and low-bandwidth communication, and real-time scene reconstruction. The framework enables efficient multi-agent 3D/4D scene synthesis through federated learning of shared diffusion model and YOLOv12 lightweight semantic extraction and local NeRF updates while maintaining privacy and scalability. The framework redesigns generative diffusion models for joint scene reconstruction, and improves cooperative scene understanding, while adding semantic-aware compression protocols. The approach can be validated through simulations and potential real-world deployment on drone testbeds, positioning it as a disruptive advancement in multi-agent AI for autonomous systems.

Due to the high flexibility and versatility, unmanned aerial vehicles (UAVs) are leveraged in various fields including surveillance and disaster rescue.However, in UAV networks, routing is vulnerable to malicious damage due to distributed topologies and high dynamics. Hence, ensuring the routing security of UAV networks is challenging. In this paper, we characterize the routing process in a time-varying UAV network with malicious nodes. Specifically, we formulate the routing problem to minimize the total delay, which is an integer linear programming and intractable to solve. Then, to tackle the network security issue, a blockchain-based trust management mechanism (BTMM) is designed to dynamically evaluate trust values and identify low-trust UAVs. To improve traditional practical Byzantine fault tolerance algorithms in the blockchain, we propose a consensus UAV update mechanism. Besides, considering the local observability, the routing problem is reformulated into a decentralized partially observable Markov decision process. Further, a multi-agent double deep Q-network based routing algorithm is designed to minimize the total delay. Finally, simulations are conducted with attacked UAVs and numerical results show that the delay of the proposed mechanism decreases by 13.39$\%$, 12.74$\%$, and 16.6$\%$ than multi-agent proximal policy optimal algorithms, multi-agent deep Q-network algorithms, and methods without BTMM, respectively.

This study aims to reinforce the theoretical foundation for diverse systems--including the axiomatic definition of intelligence--by introducing a mathematically rigorous and unified formal structure for the concept of 'state,' which has long been used without consensus or formal clarity. First, a 'hierarchical state grid' composed of two axes--state depth and mapping hierarchy--is proposed to provide a unified notational system applicable across mathematical, physical, and linguistic domains. Next, the 'Intermediate Meta-Universe (IMU)' is introduced to enable explicit descriptions of definers (ourselves) and the languages we use, thereby allowing conscious meta-level operations while avoiding self-reference and logical inconsistency. Building on this meta-theoretical foundation, this study expands inter-universal theory beyond mathematics to include linguistic translation and agent integration, introducing the conceptual division between macrocosm-inter-universal and microcosm-inter-universal operations for broader expressivity. Through these contributions, this paper presents a meta-formal logical framework--grounded in the principle of definition = state--that spans time, language, agents, and operations, providing a mathematically robust foundation applicable to the definition of intelligence, formal logic, and scientific theory at large.

Artificial intelligence (AI) systems are evolving beyond passive tools into autonomous agents capable of reasoning, adapting, and acting with minimal human intervention. Despite their growing presence, a structured framework is lacking to classify and compare these systems . This paper develops a typology of agentic AI systems, introducing eight dimensions that define their cognitive and environmental agency in an ordinal structure. Using a multi - phase methodological approach, we construct and refine this typology, which is then evaluated through a human - AI hybrid approach and further distilled into constructed types. The framework enables researchers and practitioners to analyze varying levels of agency in AI systems. By offering a structured perspective on the progression o f AI capabilities, the typology provides a foundation for assessing current systems and anticipating future developments in agentic AI.

Modern news is often comprehensive, weaving together information from diverse domains, including technology, finance, and agriculture. This very comprehensiveness creates a challenge for interpretation, as audiences typically possess specialized knowledge related to their expertise, age, or standpoint. Consequently, a reader might fully understand the financial implications of a story but fail to grasp or even actively misunderstand its legal or technological dimensions, resulting in critical comprehension gaps. In this work, we investigate how to identify these comprehension gaps and provide solutions to improve audiences' understanding of news content, particularly in the aspects of articles outside their primary domains of knowledge. We propose MADES, an agent-based framework designed to simulate societal communication. The framework utilizes diverse agents, each configured to represent a specific occupation or age group. Each agent is equipped with a memory system. These agents are then simulated to discuss the news. This process enables us to monitor and analyze their behavior and cognitive processes. Our findings indicate that the framework can identify confusions and misunderstandings within news content through its iterative discussion process. Based on these accurate identifications, the framework then designs supplementary material. We validated these outcomes using both statistical analysis and human evaluation, and the results show that agents exhibit significantly improved news understanding after receiving this supplementary material.

Learning physical interaction skills--such as dancing, handshaking, or sparring--remains a fundamental challenge for agents operating in human environments, particularly when the agent's morphology differs significantly from that of the demonstrator. Existing approaches often rely on handcrafted objectives or morphological similarity, limiting their capacity for generalization. Here, we introduce a framework that enables agents with diverse embodiments to learn whole-body interaction behaviors directly from human demonstrations. The framework extracts a compact, transferable representation of interaction dynamics--called the Embedded Interaction Graph (EIG)--which captures key spatiotemporal relationships between the interacting agents. This graph is then used as an imitation objective to train control policies in physics-based simulations, allowing the agent to generate motions that are both semantically meaningful and physically feasible. We demonstrate BuddyImitation on multiple agents, such as humans, quadrupedal robots with manipulators, or mobile manipulators and various interaction scenarios, including sparring, handshaking, rock-paper-scissors, or dancing. Our results demonstrate a promising path toward coordinated behaviors across morphologically distinct characters via cross-embodiment interaction learning. Mastering these behaviors is essential for robots, particularly non-humanoid ones, to function seamlessly in human environments. Robots may conduct daily activities such as object handover or participate in social rituals. Such interactions are critical not only for effective collaboration but also for fostering trust, acceptance, and intuitive communication between humans and machines. Beyond robotics, the ability to synthesize realistic physical interactions is also central to computer graphics and animation, where the goal is to produce lifelike and engaging character behaviors in films, games, and virtual worlds. Mastering physical interaction skills presents significant challenges due to the complex and dynamic nature of inter-agent coordination.

Constraint-based optimization is a cornerstone of robotics, enabling the design of controllers that reliably encode task and safety requirements such as collision avoidance or formation adherence. However, handcrafted constraints can fail in multi-agent settings that demand complex coordination. We introduce ReCoDe--Reinforcement-based Constraint Design--a decentralized, hybrid framework that merges the reliability of optimization-based controllers with the adaptability of multi-agent reinforcement learning. Rather than discarding expert controllers, ReCoDe improves them by learning additional, dynamic constraints that capture subtler behaviors, for example, by constraining agent movements to prevent congestion in cluttered scenarios. Through local communication, agents collectively constrain their allowed actions to coordinate more effectively under changing conditions. In this work, we focus on applications of ReCoDe to multi-agent navigation tasks requiring intricate, context-based movements and consensus, where we show that it outperforms purely handcrafted controllers, other hybrid approaches, and standard MARL baselines. We give empirical (real robot) and theoretical evidence that retaining a user-defined controller, even when it is imperfect, is more efficient than learning from scratch, especially because ReCoDe can dynamically change the degree to which it relies on this controller.

The rapid rise of Large Language Models (LLMs)-based intelligent agents underscores the need for robust, scalable evaluation frameworks. Existing methods rely on static benchmarks and labor-intensive data collection, limiting practical assessment. We introduce MCPEval, an open-source Model Context Protocol (MCP)-based framework that automates end-to-end task generation and deep evaluation of LLM agents across diverse domains. MCPEval standardizes metrics, seamlessly integrates with native agent tools, and eliminates manual effort in building evaluation pipelines. Empirical results across five real-world domains show its effectiveness in revealing nuanced, domain-specific performance. We publicly release MCPEval https://github.com/SalesforceAIResearch/MCPEval to promote reproducible and standardized LLM agent evaluation.