Gabriel Mukobi, Hannah Erlebach, Niklas Lauffer, Lewis Hammond, Alan Chan...
TL;DR通过引入一种称为 “福利外交” 的博弈规则,该研究旨在提供评估和加强合作能力的机会,以帮助研究人员发展和评估多智能体 AI 系统,并为全面社会安全做出贡献。
Abstract
The growing capabilities and increasingly widespread deployment of ai systems
necessitate robust benchmarks for measuring their cooperative capabilities.
Unfortunately, most multi-agent benchmarks are either zero