arXiv · AI· Haoyue Yang, Jingyao Li, Zhengfan Wu, Jing Liu, Xuanle Zhao, Kang Liu·· 2026-10-07AI 评分44
GameGo:基于真实资产合成轨迹训练游戏开发智能体的可扩展框架
GAMEGO: Training Game-Dev Agents with Synthetic Trajectories Anchored in Real-World Assets
AI 导读
论文提出 GameGo 框架,将简短游戏种子转化为基于行业实践的产品需求文档,并构建含 55,060 条 2D/2.5D/3D 游戏开发轨迹的 GameGoData 数据集。
来源:arXiv · AI · arxiv.org