Reproduce “Automating Structural Engineering Workflows with Large Language Model Agents” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Haoran Liang, Yufa Zhou, Mohammad Talebi Kalaleh, Qipei Mei
Researchers from the University of Alberta and Duke University developed MASSE, the first LLM-based multi-agent system designed to automate end-to-end structural engineering workflows. The system drastically reduced the time required for a racking system design task by over 98%, completing it in 2 minutes compared to 132 minutes for human engineers, while achieving high accuracy with advanced LLM backbones like o4-mini and Claude 3.5 Sonnet.
Key insights
- Multi-agent collaboration is essential for tackling the complex, long-horizon, and multi-tool nature of structural engineering problems, as single LLMs consistently fail these tasks.
- Structured, JSON-based input/output schemas and agent memory significantly enhance system reliability and performance by reducing ambiguity and preserving contextual continuity across interactions.
- Specialized reasoning LLMs (e.g., o4-mini) demonstrate superior stability and accuracy for heterogeneous engineering tasks, though a trade-off exists between performance and computational cost/runtime.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.