The primary motivation for the work is not well supported. Certainly, cities do manage thousands of intersections. While unquantified, it is not clear that the cost of training individually would surpass that of the degradation seen in the multi-env setting. These two statements seem to be conflicting. In section 5.1, the single-env results, it is not clear that FRAP is only applicable in 37 of the 112 cases.

Clarity: The paper is generally well-written, though suffers from a lack of clarity in some important sections: 4. [Equation 1] ] I believe the inner log in the right hand term of Equation (1) should not be present. I assumed it was a typo, but it is present throughout the text, even for the authors' proposed approach (e.g., in Equation 3). If intentional, why is this necessary? The paper introduces the problem scenario as a Markov game in Section 2.1; however, it introduces the notion of binary observations (which are a function of rewards here) in Section 3.1.1 This seems to suggest that perhaps the problem formulation should be corrected to a Partially Observable Markov game (POSG).

All reviewers agree this paper is a clear accept. The most critical reviewer was satisfied by the authors' rebuttal addressing his major concerns as the authors have run new experiments on a new domain, conducted a more thorough analysis of the attention mechanism in their previous experiments, and fixed some noted mistakes in their equations.

Softmax policy gradient is a popular algorithm for policy optimization in single-agent reinforcement learning, particularly since projection is not needed for each gradient update. However, in multi-agent systems, the lack of central coordination introduces significant additional difficulties in the convergence analysis. Even for a stochastic game with identical interest, there can be multiple Nash Equilibria (NEs), which disables proof techniques that rely on the existence of a unique global optimum. Moreover, the softmax parameterization introduces non-NE policies with zero gradient, making it difficult for gradient-based algorithms in seeking NEs. In this paper, we study the finite time convergence of decentralized softmax gradient play in a special form of game, Markov Potential Games (MPGs), which includes the identical interest game as a special case. We investigate both gradient play and natural gradient play, with and without \log -barrier regularization.

The paper proposes an algorithm for cooperative multi-agent games where players are trying to maximize total reward. All reviewers found the problem setting interesting and well-motivated. The two biggest concerns were the clarity of writing and how to select M when G(t) is unknown. The former was largely addressed by the authors as confirmed by the reviewers both in discussion and post-rebuttal sections of their reviews, and the scores were adjusted accordingly. The latter, however-- everyone agreed-- is problematic.

Multi-agent reinforcement learning (MARL) demonstrates significant progress in solving cooperative and competitive multi-agent problems in various environments. One of the principal challenges in MARL is the need for explicit prediction of the agents' behavior to achieve cooperation. To resolve this issue, we propose the Shared Recurrent Memory Transformer (SRMT) which extends memory transformers to multi-agent settings by pooling and globally broadcasting individual working memories, enabling agents to exchange information implicitly and coordinate their actions. We evaluate SRMT on the Partially Observable Multi-Agent Pathfinding problem in a toy Bottleneck navigation task that requires agents to pass through a narrow corridor and on a POGEMA benchmark set of tasks. In the Bottleneck task, SRMT consistently outperforms a variety of reinforcement learning baselines, especially under sparse rewards, and generalizes effectively to longer corridors than those seen during training. On POGEMA maps, including Mazes, Random, and MovingAI, SRMT is competitive with recent MARL, hybrid, and planning-based algorithms. These results suggest that incorporating shared recurrent memory into the transformerbased architectures can enhance coordination in decentralized multi-agent systems. The source code for training and evaluation is available on GitHub: https://github.com/Aloriosa/srmt. Multi-agent systems have significant potential to solve complex problems through distributed intelligence and collaboration. However, coordinating the interactions between multiple agents remains challenging, often requiring sophisticated communication protocols and decision-making mechanisms. We propose a novel approach to address this challenge by introducing a shared memory as a global workspace for agents to coordinate behavior. The global workspace theory (Baars, 1988) suggests that in the brain, there are independent functional modules that can cooperate by broadcasting information through a global workspace.

Virtual film production requires intricate decision-making processes, including scriptwriting, virtual cinematography, and precise actor positioning and actions. Motivated by recent advances in automated decision-making with language agent-based societies, this paper introduces FilmAgent, a novel LLM-based multi-agent collaborative framework for end-to-end film automation in our constructed 3D virtual spaces. FilmAgent simulates various crew roles, including directors, screenwriters, actors, and cinematographers, and covers key stages of a film production workflow: (1) idea development transforms brainstormed ideas into structured story outlines; (2) scriptwriting elaborates on dialogue and character actions for each scene; (3) cinematography determines the camera setups for each shot. A team of agents collaborates through iterative feedback and revisions, thereby verifying intermediate scripts and reducing hallucinations. We evaluate the generated videos on 15 ideas and 4 key aspects. Human evaluation shows that FilmAgent outperforms all baselines across all aspects and scores 3.98 out of 5 on average, showing the feasibility of multi-agent collaboration in filmmaking. Further analysis reveals that FilmAgent, despite using the less advanced GPT-4o model, surpasses the single-agent o1, showing the advantage of a well-coordinated multi-agent system. Lastly, we discuss the complementary strengths and weaknesses of OpenAI's text-to-video model Sora and our FilmAgent in filmmaking.

Multi-agent systems (MAS) are increasingly applied to complex task allocation in two-sided markets, where agents such as companies and customers interact dynamically. Traditional company-led Stackelberg game models, where companies set service prices, and customers respond, struggle to accommodate diverse and personalised customer demands in emerging markets like crowdsourcing. This paper proposes a customer-led Stackelberg game model for cost-efficient task allocation, where customers initiate tasks as leaders, and companies create their strategies as followers to meet these demands. We prove the existence of Nash Equilibrium for the follower game and Stackelberg Equilibrium for the leader game while discussing their uniqueness under specific conditions, ensuring cost-efficient task allocation and improved market performance. Using the satellite constellation services market as a real-world case, experimental results show a 23% reduction in customer payments and a 6.7-fold increase in company revenues, demonstrating the model's effectiveness in emerging markets.

On-ramp merging presents a critical challenge in autonomous driving, as vehicles from merging lanes need to dynamically adjust their positions and speeds while monitoring traffic on the main road to prevent collisions. To address this challenge, we propose a novel merging control scheme based on reinforcement learning, which integrates lateral control mechanisms. This approach ensures the smooth integration of vehicles from the merging lane onto the main road, optimizing both fuel efficiency and passenger comfort. Furthermore, we recognize the impact of vehicle-to-vehicle (V2V) communication on control strategies and introduce an enhanced protocol leveraging Cellular Vehicle-to-Everything (C-V2X) Mode 4. This protocol aims to reduce the Age of Information (AoI) and improve communication reliability. In our simulations, we employ two AoI-based metrics to rigorously assess the protocol's effectiveness in autonomous driving scenarios. By combining the NS3 network simulator with Python, we simulate V2V communication and vehicle control simultaneously. The results demonstrate that the enhanced C-V2X Mode 4 outperforms the standard version, while the proposed control scheme ensures safe and reliable vehicle operation during on-ramp merging.

Multi-agent systems must decide which agent is the most appropriate for a given task. We propose a novel architecture for recommending which LLM agent out of many should perform a task given a natural language prompt by extending the Sentence-BERT (SBERT) encoder model. On test data, we are able to achieve a top-1 accuracy of 92.2% with each classification taking less than 300 milliseconds. In contrast to traditional classification methods, our architecture is computationally cheap, adaptive to new classes, interpretable, and controllable with arbitrary metrics through reinforcement learning. By encoding natural language prompts into sentence embeddings, our model captures the semantic content relevant to recommending an agent. The distance between sentence embeddings that belong to the same agent is then minimized through fine-tuning and aligned to human values through reinforcement learning from human feedback. This allows the classification of natural language prompts based on their nearest neighbors by measuring the cosine similarity between embeddings. This work is made possible through the generation of a synthetic dataset for agent recommendation, which we have open-sourced to the public along with the code for AgentRec recommendation system at https://github.com/joshprk/agentrec.