metadata
title: AEC-Bench
sdk: static
AEC-Bench
AEC-Bench is an open benchmark and Python toolkit for evaluating agentic AI systems on realistic Architecture, Engineering, and Construction tasks.
The project combines generated engineering tasks, executable verifiers, model rollout ledgers, and trace artifacts so evaluation can be inspected beyond a single leaderboard score: by task family, difficulty, information visibility, tool use, cost, and failure mode.
- Paper: AEC-Bench: A Multimodal Benchmark for Agentic Systems in Architecture, Engineering, and Construction
- Code: TheodoreGalanos/aec-bench
- First release dataset: aec-bench/release-model-rollouts
This organisation hosts datasets, rollout artifacts, and benchmark releases for the AEC-Bench project.