Real-World Construction Documents for Multimodal AI Benchmarking

AEC-Bench is a multimodal collection of real-world Architecture, Engineering, and Construction documents, including construction drawings and floor plans. The dataset was created by nomic-ai and was last updated in April 2026. It is structured for benchmarking tasks across scopes and task families.

Use Cases

• Train models for visual question answering using images of construction drawings and associated text queries.
• Benchmark multimodal reasoning on floor plans by analyzing spatial layouts and textual annotations.
• Develop document understanding systems that process both the visual modality of engineering diagrams and the text modality of associated labels and descriptions.

Strengths

• Dataset is explicitly designed for benchmarking with a defined task taxonomy covering multiple scopes and families.
• Contains real-world documents from the Architecture, Engineering, and Construction domain, providing domain-specific data.

Limitations

• Specific quantitative details like row count, file size, and column structure are not provided in the input.
• The geographic origin and temporal range of the source documents are unknown, which may limit generalizability assessments.

Provenance

Source

nomic-ai via Hugging Face.

Collection Method

Collection of real-world AEC documents; annotations are expert-generated.

Time Range

null

Freshness

null

Geography

null