arXiv · Artificial Intelligence· Haoyue Yang, Jingyao Li, Zhengfan Wu, Jing Liu, Xuanle Zhao, Kang Liu·· 4 hr agoAI score34
GameGo：用锚定真实资产的合成轨迹训练游戏开发 Agent
GAMEGO: Training Game-Dev Agents with Synthetic Trajectories Anchored in Real-World Assets
AI brief
GameGo 提出一个可扩展框架，把简短游戏种子转化为基于行业游戏开发实践的完整 Product Requirements Documents，并用任务特定动态压缩在保留核心玩法约束的同时不限制设计探索。
Source: arXiv · Artificial Intelligence · arxiv.org