VibeGame:基于AI原生引擎与自进化对抗智能体团队的提示到游戏开发
原标题:VibeGame: Prompt-to-Game Development with AI-Native Engine and Self-Evolving Adversarial Agent Team
AI 摘要
南京大学、南洋理工大学和Liblib.ai的研究者提出VibeGame框架,通过自然语言提示和可选参考图像生成完整可玩的2D游戏项目。该框架包含一个AI原生游戏引擎,将项目状态表示为结构化文本并支持帧同步运行时控制,同时设计了由八个专用智能体组成的对抗性智能体团队,分离实现与审查角色并通过多轮循环修复问题。在GameCraft-Bench子集上,VibeGame在P2G设置下持续优于包括GPT-6.1-Sol在内的多种智能体框架。
正文节选
promptfile 1]PRLab, Nanjing University 2]S-Lab, Nanyang Technological University 3]Liblib.ai \contribution[*]Equal Contribution \contribution[†]Corresponding Author \checkdata[Project Page]https://vibegame.tettet.org \checkdata[Code]https://github.com/tettethu/VibeGame \correspondencechenyang.si@nju.edu.cn \teaser VibeGame: Prompt-to-Game Development with AI-Native Engine and Self-Evolving Adversarial Agent Team Abstract Large language model agents have enabled software development through natur