ROAR：实时约束下多人射击游戏中的LMM智能体基准
原标题：ROAR: Benchmarking LMM Agents under Real-Time Constraints in Multiplayer Action-Shooter Games
AI 摘要
研究团队提出 ROAR，一个多人动作射击游戏环境，用于在实时约束下评测基于大型多模态模型（LMM）的智能体。现有基准多为回合制，推理时环境与其他智能体暂停，掩盖了实时决策挑战。ROAR 采用多智能体异步 Gym 接口，将环境与各智能体分线程运行，支持合作、竞争、混合及单智能体场景，并提供多模态观测与轻量部署。
正文节选
ROAR: Benchmarking LMM Agents under Real-Time Constraints in Multiplayer Action-Shooter Games Abstract Large multimodal models (LMMs) are increasingly deployed as agents that perceive, reason, and act in interactive multi-agent environments. Yet most existing benchmarks for LMM-based agents in multi-agent settings remain turn-based: the environment and other agents pause while an agent performs inference. Such protocols obscure a fundamental challenge of real-time constraints, where the world co