Multi-agent collaboration enhances the perception capabilities of individual agents through information sharing. However, in real-world applications, differences in sensors and models across heterogeneous agents inevitably lead to domain gaps during collaboration. Existing approaches based on adaptation and reconstruction fail to support pragmatic heterogeneous collaboration due to two key limitations: (1) Intrusive retraining of the encoder or core modules disrupts the established semantic consistency among agents; and (2) accommodating new agents incurs high computational costs, limiting scalability. To address these challenges, we present a novel Generative Communication mechanism (GenComm) that facilitates seamless perception across heterogeneous multi-agent systems through feature generation, without altering the original network, and employs lightweight numerical alignment of spatial information to efficiently integrate new agents at minimal cost. Specifically, a tailored Deformable Message Extractor is designed to extract spatial message for each collaborator, which is then transmitted in place of intermediate features. The Spatial-Aware Feature Generator, utilizing a conditional diffusion model, generates features aligned with the ego agent's semantic space while preserving the spatial information of the collaborators. These generated features are further refined by a Channel Enhancer before fusion. Experiments conducted on the OPV2V-H, DAIR-V2X and V2X-Real datasets demonstrate that GenComm outperforms existing state-of-the-art methods, achieving an 81% reduction in both computational cost and parameter count when incorporating new agents.

Humans are remarkably adept at collaboration, able to infer the strengths and weaknesses of new partners in order to work successfully towards shared goals. To build AI systems with this capability, we must first understand its building blocks: does such flexibility require explicit, dedicated mechanisms for modelling others--or can it emerge spontaneously from the pressures of open-ended cooperative interaction? To investigate this question, we train simple model-free RNN agents to collaborate with a population of diverse partners. Using the'Overcooked-AI' environment, we collect data from thousands of collaborative teams, and analyse agents' internal hidden states. Despite a lack of additional architectural features, inductive biases, or auxiliary objectives, the agents nevertheless develop structured internal representations of their partners' task abilities, enabling rapid adaptation and generalisation to novel collaborators. We investigated these internal models through probing techniques, and large-scale behavioural analysis. Notably, we find that structured partner modelling emerges when agents can influence partner behaviour by controlling task allocation. Our results show that partner modelling can arise spontaneously in model-free agents--but only under environmental conditions that impose the right kind of social pressure.

The significant difference between the deployment-time partners' distribution and the training partners' distribution determined by the

The goal of zero-shot human-AI coordination is to develop an agent capable of collaborating with humans without relying on human data.

Uncertainty-aware prediction is essential for safe motion planning, especially when using learned models to forecast the behavior of surrounding agents. Conformal prediction is a statistical tool often used to produce uncertainty-aware prediction regions for machine learning models. Most existing frameworks utilizing conformal prediction-based uncertainty predictions assume that the surrounding agents are non-interactive. This is because in closed-loop, as uncertainty-aware agents change their behavior to account for prediction uncertainty, the surrounding agents respond to this change, leading to a distribution shift which we call endogenous distribution shift. To address this challenge, we introduce an iterative conformal prediction framework that systematically adapts the uncertainty-aware ego-agent controller to the endogenous distribution shift. The proposed method provides probabilistic safety guarantees while adapting to the evolving behavior of reactive, non-ego agents. We establish a model for the endogenous distribution shift and provide the conditions for the iterative conformal prediction pipeline to converge under such a distribution shift. We validate our framework in simulation for 2- and 3- agent interaction scenarios, demonstrating collision avoidance without resulting in overly conservative behavior and an overall improvement in success rates of up to 9.6% compared to other conformal prediction-based baselines.

Learning to collaborate with previously unseen partners is a fundamental generalization challenge in multi-agent learning, known as Ad Hoc Teamwork (AHT). Existing AHT approaches often adopt a two-stage pipeline, where first, a fixed population of teammates is generated with the idea that they should be representative of the teammates that will be seen at deployment time, and second, an AHT agent is trained to collaborate well with agents in the population. To date, the research community has focused on designing separate algorithms for each stage. This separation has led to algorithms that generate teammates with limited coverage of possible behaviors, and that ignore whether the generated teammates are easy to learn from for the AHT agent. Furthermore, algorithms for training AHT agents typically treat the set of training teammates as static, thus attempting to generalize to previously unseen partner agents without assuming any control over the set of training teammates. This paper presents a unified framework for AHT by reformulating the problem as an open-ended learning process between an AHT agent and an adversarial teammate generator. We introduce ROTATE, a regret-driven, open-ended training algorithm that alternates between improving the AHT agent and generating teammates that probe its deficiencies. Experiments across diverse two-player environments demonstrate that ROTATE significantly outperforms baselines at generalizing to an unseen set of evaluation teammates, thus establishing a new standard for robust and generalizable teamwork.

While game-theoretic planning frameworks are effective at modeling multi-agent interactions, they require solving large optimization problems where the number of variables increases with the number of agents, resulting in long computation times that limit their use in large-scale, real-time systems. To address this issue, we propose 1) PSN Game: a learning-based, game-theoretic prediction and planning framework that reduces runtime by learning a Player Selection Network (PSN); and 2) a Goal Inference Network (GIN) that makes it possible to use the PSN in incomplete information games where agents' intentions are unknown. A PSN outputs a player selection mask that distinguishes influential players from less relevant ones, enabling the ego player to solve a smaller, masked game involving only selected players. By reducing the number of players in the game, and therefore reducing the number of variables in the corresponding optimization problem, PSN directly lowers computation time. The PSN Game framework is more flexible than existing player selection methods as it 1) relies solely on observations of players' past trajectories, without requiring full state, action, or other game-specific information; and 2) requires no online parameter tuning. Experiments in both simulated scenarios and human trajectory datasets demonstrate that PSNs outperform baseline selection methods in 1) prediction accuracy; and 2) planning safety. PSNs also generalize effectively to real-world scenarios in which agents' objectives are unknown without fine-tuning. By selecting only the most relevant players for decision-making, PSN Game offers a general mechanism for reducing planning complexity that can be seamlessly integrated into existing multi-agent planning frameworks.

The significant difference between the deployment-time partners' distribution and the training partners' distribution determined by the