Autonomous driving faces critical challenges in rare long-tail events and complex multi-agent interactions, which are scarce in real-world data yet essential for robust safety validation. This paper presents a high-fidelity scenario generation framework that integrates a conditional variational autoencoder (CVAE) with a large language model (LLM). The CVAE encodes historical trajectories and map information from large-scale naturalistic datasets to learn latent traffic structures, enabling the generation of physically consistent base scenarios. This knowledge-driven optimization balances realism with controllability, ensuring that generated scenarios remain both plausible and risk-sensitive. Extensive experiments in CARLA and SMARTS demonstrate that our framework substantially increases the coverage of high-risk and long-tail events, improves consistency between simulated and real-world traffic distributions, and exposes autonomous driving systems to interactions that are significantly more challenging than those produced by existing rule-or data-driven methods. These results establish a new pathway for safety validation, enabling principled stress-testing of autonomous systems under rare but consequential events. Introduction The safety and reliability of autonomous driving depend on rigorous validation under diverse test conditions, especially in high-risk, highly interactive, and safety-critical scenarios (Wang et al., 2021; Hossain, 2025). Yet such events are extremely scarce in real-world datasets, creating a persistent gap between development testing and deployment needs. Simulation-based methods provide an effective alternative by generating large numbers of rare and adversarial environments, thereby alleviating data scarcity and enabling controlled safety evaluation (Huang et al., 2020). To address these challenges, this paper proposes a risk knowledge-guided traffic scene generation framework that integrates a Conditional Variational Autoencoder (CV AE) with a Large Language Model (LLM). Unlike prior works that merely sample or replay specific risky cases, the proposed framework establishes a general and controllable pipeline for synthesizing diverse safety-critical scenarios under varying risk conditions. The CVAE learns latent spatiotemporal representations from real-world trajectories and maps to generate physically coherent base scenes, while the LLM acts as a knowledge-driven controller that interprets scene semantics, analyzes multi-agent risk interactions, and dynamically adjusts optimization objectives to guide the generation toward desired levels of behavioral complexity and risk exposure.

One of the drawbacks of INs is scaling with the number of interactions in the system (typically quadratic or higher order in the number of agents). In this paper we introduce V AIN, a novel attentional architecture for multi-agent predictive modeling that scales linearly with the number of agents. We show that V AIN is effective for multi-agent predictive modeling.

Large Language Model (LLM)-based multi-agent systems are increasingly used to simulate human interactions and solve collaborative tasks. A common practice is to assign agents with personas to encourage behavioral diversity. However, this raises a critical yet underexplored question: do personas introduce biases into multi-agent interactions? This paper presents a systematic investigation into persona-induced biases in multi-agent interactions, with a focus on social traits like trustworthiness (how an agent's opinion is received by others) and insistence (how strongly an agent advocates for its opinion). Through a series of controlled experiments in collaborative problem-solving and persuasion tasks, we reveal that (1) LLM-based agents exhibit biases in both trustworthiness and insistence, with personas from historically advantaged groups (e.g., men and White individuals) perceived as less trustworthy and demonstrating less insistence; and (2) agents exhibit significant in-group favoritism, showing a higher tendency to conform to others who share the same persona. These biases persist across various LLMs, group sizes, and numbers of interaction rounds, highlighting an urgent need for awareness and mitigation to ensure the fairness and reliability of multi-agent systems.

While stereotypes are well-documented in human social interactions, AI systems are often presumed to be less susceptible to such biases. Previous studies have focused on biases inherited from training data, but whether stereotypes can emerge spontaneously in AI agent interactions merits further exploration. Through a novel experimental framework simulating workplace interactions with neutral initial conditions, we investigate the emergence and evolution of stereotypes in LLM-based multi-agent systems. Our findings reveal that (1) LLM-Based AI agents develop stereotype-driven biases in their interactions despite beginning without predefined biases; (2) stereotype effects intensify with increased interaction rounds and decision-making power, particularly after introducing hierarchical structures; (3) these systems exhibit group effects analogous to human social behavior, including halo effects, confirmation bias, and role congruity; and (4) these stereotype patterns manifest consistently across different LLM architectures. Through comprehensive quantitative analysis, these findings suggest that stereotype formation in AI systems may arise as an emergent property of multi-agent interactions, rather than merely from training data biases. Our work underscores the need for future research to explore the underlying mechanisms of this phenomenon and develop strategies to mitigate its ethical impacts.

Multi-agent systems, which consist of multiple AI models interacting within a shared environment, are increasingly used for persona-based interactions. However, if not carefully designed, these systems can reinforce implicit biases in large language models (LLMs), raising concerns about fairness and equitable representation. We present MALIBU, a novel benchmark developed to assess the degree to which LLM-based multi-agent systems implicitly reinforce social biases and stereotypes. MALIBU evaluates bias in LLM-based multi-agent systems through scenario-based assessments. AI models complete tasks within predefined contexts, and their responses undergo evaluation by an LLM-based multi-agent judging system in two phases. In the first phase, judges score responses labeled with specific demographic personas (e.g., gender, race, religion) across four metrics. In the second phase, judges compare paired responses assigned to different personas, scoring them and selecting the superior response. Our study quantifies biases in LLM-generated outputs, revealing that bias mitigation may favor marginalized personas over true neutrality, emphasizing the need for nuanced detection, balanced fairness strategies, and transparent evaluation benchmarks in multi-agent systems.

Abstract-- From autonomous driving to package delivery, ensuring safe yet efficient multi-agent interaction is challenging as the interaction dynamics are influenced by hard-to-model factors such as social norms and contextual cues. Understanding these influences can aid in the design and evaluation of sociallyaware autonomous agents whose behaviors are aligned with human values. In this work, we seek to codify factors governing safe multi-agent interactions via the lens of responsibility, i.e., an agent's willingness to deviate from their desired control to accommodate safe interaction with others. Specifically, we propose a data-driven modeling approach based on control barrier functions and differentiable optimization that efficiently learns agents' responsibility allocation from data. We demonstrate on synthetic and real-world datasets that we can obtain Figure 1: In a) and b), two cars are swapping lanes on a highway, but an interpretable and quantitative understanding of how much their desired controls lead to collision. In c) and d), we see how the agents adjust their behavior to ensure the safety of others given agents may deviate from their ideal trajectories, according to two their current environment.

Neural Information Processing Systems http://nips.cc/

Understanding collective behavior and how it evolves is important to ensure that robot swarms can be trusted in a shared environment. One way to understand the behavior of the swarm is through collective behavior reconstruction using prior demonstrations. Existing approaches often require access to the swarm controller which may not be available. We reconstruct collective behaviors in distinct swarm scenarios involving shared environments without using swarm controller information. We achieve this by transforming prior demonstrations into features that sufficiently describe multi-agent interactions before behavior reconstruction with multi-agent generative adversarial imitation learning (MA-GAIL). We show that our approach outperforms existing algorithms in all investigated swarm scenarios, and can be used to observe and reconstruct a swarm's behavior for further analysis and testing, which might be impractical or undesirable on the original robot swarm.

In interactive multi-agent settings, decision-making complexity arises from agents' interconnected objectives. Dynamic game theory offers a formal framework for analyzing such intricacies. Yet, solving dynamic games and determining Nash equilibria pose computational challenges due to the need of solving coupled optimal control problems. To address this, our key idea is to leverage potential games, which are games with a potential function that allows for the computation of Nash equilibria by optimizing the potential function. We argue that dynamic potential games, can effectively facilitate interactive decision-making in many multi-agent interactions. We will identify structures in realistic multi-agent interactive scenarios that can be transformed into weighted potential dynamic games. We will show that the open-loop Nash equilibria of the resulting weighted potential dynamic game can be obtained by solving a single optimal control problem. We will demonstrate the effectiveness of the proposed method through various simulation studies, showing close proximity to feedback Nash equilibria and significant improvements in solve time compared to state-of-the-art game solvers.

In this work, we develop a scalable, local trajectory optimization algorithm that enables robots to interact with other robots. It has been shown that agents' interactions can be successfully captured in game-theoretic formulations, where the interaction outcome can be best modeled via the equilibria of the underlying dynamic game. However, it is typically challenging to compute equilibria of dynamic games as it involves simultaneously solving a set of coupled optimal control problems. Existing solvers operate in a centralized fashion and do not scale up tractably to multiple interacting agents. We enable scalable distributed game-theoretic planning by leveraging the structure inherent in multi-agent interactions, namely, interactions belonging to the class of dynamic potential games. Since equilibria of dynamic potential games can be found by minimizing a single potential function, we can apply distributed and decentralized control techniques to seek equilibria of multi-agent interactions in a scalable and distributed manner. We compare the performance of our algorithm with a centralized interactive planner in a number of simulation studies and demonstrate that our algorithm results in better efficiency and scalability. We further evaluate our method in hardware experiments involving multiple quadcopters.