Wuli-Shili-Renli-Jili System Approach and Its Application in Large Model Value Alignment

Jiayin Qi, Kun Jia, Yuxin Zhang, Xiaobing Feng, Haiyan Yan

Abstract: As artificial intelligence systems enter a rapid development phase, large models, serving as a new type of hybrid intelligence system for human-machine collaboration, still lack a system methodology that suits them. This study extends the Wuli-Shili-Renli (WSR) system approach, innovatively proposing a Wuli- Shili- Renli- Jili (WSRJ) system approach suitable for large models. It constructs the connotation of the Jili element and explores the application of WSRJ system approach in the technology research of large model value alignment, providing a multi-dimensional system analysis framework of function, logic, humanity, and technology.