Multi-agent systems (MAS) and reinforcement learning (RL) are widely used to enhance the agentic capabilities of large language models (LLMs). MAS improves task performance through role-based orchestration, while RL uses environmental rewards to learn stronger policies, such as GRPO-style optimization. However, applying on-policy RL to MAS remains underexplored and presents unique challenges. Algorithmically, standard GRPO grouping assumptions break down because prompts vary by role and by turn. System-wise, the training stack must support MAS-workflow rollouts and on-policy updates for both single-policy and multi-policy models. We propose AT-GRPO, which includes (i) an agent- and turn-wise grouped RL algorithm tailored to MAS and (ii) a training system that supports both single- and multi-policy regimes. Across game, planning, coding, and math tasks, AT-GRPO delivers substantial gains. On long-horizon planning, it increases accuracy from a 14.0 to 47.0 percent single-agent RL baseline to 96.0 to 99.5 percent. It also improves reasoning performance, with average gains of 3.87 to 7.62 percent on coding tasks and 9.0 to 17.93 percent on math. Code and environments are available at: https://github.com/pettingllms-ai/PettingLLMs.

翻译：多智能体系统（MAS）与强化学习（RL）被广泛用于增强大语言模型（LLMs）的智能体能力。MAS通过基于角色的编排提升任务性能，而RL则利用环境奖励学习更优策略，例如GRPO风格的优化方法。然而，将同策略RL应用于MAS的研究仍显不足，且面临独特挑战。在算法层面，标准GRPO的分组假设因提示信息随角色和轮次变化而失效。在系统层面，训练框架需支持MAS工作流的轨迹采样，并同时适用于单策略与多策略模型的同策略更新。我们提出了AT-GRPO方法，其包含：（i）专为MAS设计的、按智能体与轮次分组的RL算法；（ii）支持单策略与多策略模式的训练系统。在游戏、规划、编程和数学任务上的实验表明，AT-GRPO带来了显著性能提升。在长程规划任务中，其准确率将从单智能体RL基线的14.0%至47.0%提升至96.0%至99.5%。该方法同时提升了推理性能，在编程任务上平均增益达3.87%至7.62%，在数学任务上平均增益达9.0%至17.93%。代码与环境已开源：https://github.com/pettingllms-ai/PettingLLMs。