GameCraft-Bench evaluates agents on building playable games end-to-end

Arxiv CS.CL (NLP/language models)

The benchmark requires coding agents to transform natural-language specifications into playable games within a real game engine. It tests end-to-end game generation, from specification to interactive system, without supervised game data.

See 1 more source

GameCraft-Bench: Can Agents Build Playable Games End-to-End in a Real Game Engine?5 hours agopmttyjiDiscuss

Arxiv CS.CL (NLP/language models)·Tongxu LuoRongsheng WangJiaxi BiChenming XuZhengyang TangJianlong ChenJuhao LiangKe JiShuqi GuoYuhao DuFan BuWenyu DuXiaotong ZhangKyle LiShaobo WangLinfeng ZhangYuxuan LiuXin LaiChenxin LiYiduo GuoZhexin ZhangXiny

1 day ago