Abstract: The architectural industry produces exten- sive documents, including method state- ments—expository documents that integrate multi-source data into actionable guidance. Manual drafting however is labor-intensive and time-consuming. This paper introduces Archi- DocGen, a multi-agent framework automating method statement generation. Unlike tradi- tional approaches relying on static templates or single-pass generation, ArchiDocGen decom- poses the task into three steps: outline gener- ation, section-based content generation, and polishing, each handled by specialized agents. To provide domain expertise, ArchiDocGen employs a section-based retriever to fetch and synthesize relevant documents from its custom knowledge base. Each section is generated through iterative reasoning of a section-based chain-of-thought (SeCoT) scheme, followed by refinement to meet professional standards. To evaluate the generated method statements, we partner with the industry to establish a multi- dimensional evaluation system by combining automatic and empirical methods. Experiments show that ArchiDocGen achieves 4.38 Con- tentScore, excelling in specialization, complete- ness, organization, and clarity. Additionally, a web-based application for ArchiDocGen is de- veloped and deployed with industry partners.
Loading