Accurate prediction of future agent trajectories is a critical challenge for ensuring safe and efficient autonomous navigation, particularly in complex urban environments characterized by multiple plausible future scenarios. In this paper, we present a novel hybrid approach that integrates learning-based with physics-based constraints to address the multi-modality inherent in trajectory prediction. Our method employs a variational Bayesian mixture model to effectively capture the diverse range of potential future behaviors, moving beyond traditional unimodal assumptions. Unlike prior approaches that predominantly treat trajectory prediction as a data-driven regression task, our framework incorporates physical realism through sector-specific boundary conditions and Model Predictive Control (MPC)-based smoothing. These constraints ensure that predicted trajectories are not only data-consistent but also physically plausible, adhering to kinematic and dynamic principles. Furthermore, our method produces interpretable and diverse trajectory predictions, enabling enhanced downstream decision-making and planning in autonomous driving systems. We evaluate our approach on two benchmark datasets, demonstrating superior performance compared to existing methods. Comprehensive ablation studies validate the contributions of each component and highlight their synergistic impact on prediction accuracy and reliability. By balancing data-driven insights with physics-informed constraints, our approach offers a robust and scalable solution for navigating the uncertainties of real-world urban environments.

Nearly all human work is collaborative; thus, the evaluation of real-world NLP applications often requires multiple dimensions that align with diverse human perspectives. As real human evaluator resources are often scarce and costly, the emerging "LLM-as-a-judge" paradigm sheds light on a promising approach to leverage LLM agents to believably simulate human evaluators. Yet, to date, existing LLM-as-a-judge approaches face two limitations: persona descriptions of agents are often arbitrarily designed, and the frameworks are not generalizable to other tasks. To address these challenges, we propose MAJ-EVAL, a Multi-Agent-as-Judge evaluation framework that can automatically construct multiple evaluator personas with distinct dimensions from relevant text documents (e.g., research papers), instantiate LLM agents with the personas, and engage in-group debates with multi-agents to Generate multi-dimensional feedback. Our evaluation experiments in both the educational and medical domains demonstrate that MAJ-EVAL can generate evaluation results that better align with human experts' ratings compared with conventional automated evaluation metrics and existing LLM-as-a-judge methods.

--Recent advances in autonomous vehicle (A V) behavior planning have shown impressive social interaction capabilities when interacting with other road users. However, achieving human-like prediction and decision-making in interactions with vulnerable road users remains a key challenge in complex multi-agent interactive environments. Existing research focuses primarily on crowd navigation for small mobile robots, which cannot be directly applied to A Vs due to inherent differences in their decision-making strategies and dynamic boundaries. Moreover, pedestrians in these multi-agent simulations follow fixed behavior patterns that cannot dynamically respond to A V actions. T o overcome these limitations, this paper proposes a novel framework for modeling interactions between the A V and multiple pedestrians. In this framework, a cognitive process modeling approach inspired by the Free Energy Principle is integrated into both the A V and pedestrian models to simulate more realistic interaction dynamics. Specifically, the proposed pedestrian Cognitive-Risk Social Force Model adjusts goal-directed and repulsive forces using a fused measure of cognitive uncertainty and physical risk to produce human-like trajectories. Meanwhile, the A V leverages this fused risk to construct a dynamic, risk-aware adjacency matrix for a Graph Convolutional Network within a Soft Actor-Critic architecture, allowing it to make more reasonable and informed decisions. Simulation results indicate that our proposed framework effectively improves safety, efficiency, and smoothness of A V navigation compared to the state-of-the-art method. N recent years, rapid advancements in autonomous driving technology have enabled autonomous vehicles (A V) to expand beyond simple, structured highway environments and to be increasingly deployed in more complex urban environments [1]. They are expected to become a key component of future urban transportation systems [2]. Unlike structured roads with clear traffic rules and physical lane separations, shared spaces in cities such as squares, campuses, and residential areas usually lack right-of-way regulations and explicit physical boundaries [3] between vehicles and pedestrians. Y afei Wang is with the School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China (email: wyfjlu@sjtu.edu.cn) The high uncertainty and dynamic nature of human behavior [5], especially when multiple pedestrians are moving at the same time, significantly complicate the A V's decision-making process. To ensure both safety and efficiency, A Vs must make real-time decisions and continuously adapt their strategies in response to surrounding pedestrian behaviors.

Situationally-aware artificial agents operating with competence in natural environments face several challenges: spatial awareness, object affordance detection, dynamic changes and unpredictability. A critical challenge is the agent's ability to identify and monitor environmental elements pertinent to its objectives. Our research introduces a neurosymbolic modular architecture for reactive robotics. Our system combines a neural component performing object recognition over the environment and image processing techniques such as optical flow, with symbolic representation and reasoning. The reasoning system is grounded in the embodied cognition paradigm, via integrating image schematic knowledge in an ontological structure. The ontology is operatively used to create queries for the perception system, decide on actions, and infer entities' capabilities derived from perceptual data. The combination of reasoning and image processing allows the agent to focus its perception for normal operation as well as discover new concepts for parts of objects involved in particular interactions. The discovered concepts allow the robot to autonomously acquire training data and adjust its subsymbolic perception to recognize the parts, as well as making planning for more complex tasks feasible by focusing search on those relevant object parts. We demonstrate our approach in a simulated world, in which an agent learns to recognize parts of objects involved in support relations. While the agent has no concept of handle initially, by observing examples of supported objects hanging from a hook it learns to recognize the parts involved in establishing support and becomes able to plan the establishment/destruction of the support relation. This underscores the agent's capability to expand its knowledge through observation in a systematic way, and illustrates the potential of combining deep reasoning [...].

This paper preliminarily investigates the duality between flow matching in generative models and particle swarm optimization (PSO) in evolutionary computation. Through theoretical analysis, we reveal the intrinsic connections between these two approaches in terms of their mathematical formulations and optimization mechanisms: the vector field learning in flow matching shares similar mathematical expressions with the velocity update rules in PSO; both methods follow the fundamental framework of progressive evolution from initial to target distributions; and both can be formulated as dynamical systems governed by ordinary differential equations. Our study demonstrates that flow matching can be viewed as a continuous generalization of PSO, while PSO provides a discrete implementation of swarm intelligence principles. This duality understanding establishes a theoretical foundation for developing novel hybrid algorithms and creates a unified framework for analyzing both methods. Although this paper only presents preliminary discussions, the revealed correspondences suggest several promising research directions, including improving swarm intelligence algorithms based on flow matching principles and enhancing generative models using swarm intelligence concepts.

Aligning AI systems with human values and the value-based preferences of various stakeholders (their value systems) is key in ethical AI. In value-aware AI systems, decision-making draws upon explicit computational representations of individual values (groundings) and their aggregation into value systems. As these are notoriously difficult to elicit and calibrate manually, value learning approaches aim to automatically derive computational models of an agent's values and value system from demonstrations of human behaviour. Nonetheless, social science and humanities literature suggest that it is more adequate to conceive the value system of a society as a set of value systems of different groups, rather than as the simple aggregation of individual value systems. Accordingly, here we formalize the problem of learning the value systems of societies and propose a method to address it based on heuristic deep clustering. The method learns socially shared value groundings and a set of diverse value systems representing a given society by observing qualitative value-based preferences from a sample of agents. We evaluate the proposal in a use case with real data about travelling decisions.

Personal use of this material is permitted. Abstract --Understanding and predicting player movement in multiplayer games is crucial for achieving use cases such as player-mimicking bot navigation, preemptive bot control, strategy recommendation, and real-time player behavior analytics. However, the complex environments allow for a high degree of navigational freedom, and the interactions and team-play between players require models that make effective use of the available heterogeneous input data. This paper presents a multimodal architecture for predicting future player locations on a dynamic time horizon, using a U-Net -based approach for calculating endpoint location probability heatmaps, conditioned using a multimodal feature encoder . The application of a multi-head attention mechanism for different groups of features allows for communication between agents. In doing so, the architecture makes efficient use of the multimodal game state including image inputs, numerical and categorical features, as well as dynamic game data. Consequently, the presented technique lays the foundation for various downstream tasks that rely on future player positions such as the creation of player-predictive bot behavior or player anomaly detection. Predicting the future position of players in team-based video game environments is important for a variety of tasks such as AI-based decision-making, strategy optimization of bots, and real-time player behavior analysis. However, such predictions can be challenging, especially in dynamic game environments with complex interactions between multiple entities. Besides applications in video games, the task of location prediction has found wide-ranging interest in domains such as robotics [1], autonomous driving [2], and sports analytics [3].

Autonomy is a double-edged sword for AI agents, simultaneously unlocking transformative possibilities and serious risks. How can agent developers calibrate the appropriate levels of autonomy at which their agents should operate? We argue that an agent's level of autonomy can be treated as a deliberate design decision, separate from its capability and operational environment. In this work, we define five levels of escalating agent autonomy, characterized by the roles a user can take when interacting with an agent: operator, collaborator, consultant, approver, and observer. Within each level, we describe the ways by which a user can exert control over the agent and open questions for how to design the nature of user-agent interaction. We then highlight a potential application of our framework towards AI autonomy certificates to govern agent behavior in single- and multi-agent systems. We conclude by proposing early ideas for evaluating agents' autonomy. Our work aims to contribute meaningful, practical steps towards responsibly deployed and useful AI agents in the real world.

Recent advances have enabled LLM-powered AI agents to autonomously execute complex tasks by combining language model reasoning with tools, memory, and web access. But can these systems be trusted to follow deployment policies in realistic environments, especially under attack? To investigate, we ran the largest public red-teaming competition to date, targeting 22 frontier AI agents across 44 realistic deployment scenarios. Participants submitted 1.8 million prompt-injection attacks, with over 60,000 successfully eliciting policy violations such as unauthorized data access, illicit financial actions, and regulatory noncompliance. We use these results to build the Agent Red Teaming (ART) benchmark - a curated set of high-impact attacks - and evaluate it across 19 state-of-the-art models. Nearly all agents exhibit policy violations for most behaviors within 10-100 queries, with high attack transferability across models and tasks. Importantly, we find limited correlation between agent robustness and model size, capability, or inference-time compute, suggesting that additional defenses are needed against adversarial misuse. Our findings highlight critical and persistent vulnerabilities in today's AI agents. By releasing the ART benchmark and accompanying evaluation framework, we aim to support more rigorous security assessment and drive progress toward safer agent deployment.

The rapid growth of crypto markets has opened new opportunities for investors, but at the same time exposed them to high volatility. To address the challenge of managing dynamic portfolios in such an environment, this paper presents a practical application of a multi-agent system designed to autonomously construct and evaluate crypto-asset allocations. Using data on daily frequencies of the ten most capitalized cryptocurrencies from 2020 to 2025, we compare two automated investment strategies. These are a static equal weighting strategy and a rolling-window optimization strategy, both implemented to maximize the evaluation metrics of the Modern Portfolio Theory (MPT), such as Expected Return, Sharpe and Sortino ratios, while minimizing volatility. Each step of the process is handled by dedicated agents, integrated through a collaborative architecture in Crew AI. The results show that the dynamic optimization strategy achieves significantly better performance in terms of risk-adjusted returns, both in-sample and out-of-sample. This highlights the benefits of adaptive techniques in portfolio management, particularly in volatile markets such as cryptocurrency markets. The following methodology proposed also demonstrates how multi-agent systems can provide scalable, auditable, and flexible solutions in financial automation.