DeskCraft benchmark evaluates desktop agents on professional workflows

Arxiv CS.AI (top papers)

DeskCraft benchmarks desktop agents on long-horizon professional tasks in creative and engineering software. It emphasizes human-in-the-loop collaboration where agents must proactively seek information and users provide additional context.

benchmarks

Arxiv CS.AI (top papers)·Wenkai WangTao XiongJingchen NiYunpeng BaoXiyun LiTianqi LiuHongcan GuoZilong HuangShengyu Zhang

8 days ago