Ishigaki-IDS-Bench: Bilingual Building Requirements to IDS XML

Ishigaki-IDS-Bench is a bilingual benchmark for generating buildingSMART IDS 1.0 XML from user-facing building requirements. The dataset was created by ONESTRUCTION and last updated on Hugging Face in May 2026. Each data row is formatted as a conversational exchange between a user and an assistant, covering the architectural domain.

Use Cases

• Fine-tuning language models for Information Delivery Specification (IDS) generation based on the described bilingual benchmark task.
• Evaluating model performance on converting natural language building requirements to structured XML based on the buildingSMART IDS 1.0 standard.
• Training conversational AI assistants for the AEC domain based on the single-turn, user-assistant message format.
• Researching cross-lingual transfer learning in technical domains based on the bilingual (Japanese and likely English) nature of the data.

Strengths

• Dataset is explicitly designed as a benchmark for a specific technical standard (buildingSMART IDS 1.0).
• Data is structured in a standardized conversational format suitable for training language models.
• Covers the architectural domain with potential support for multiple IFC versions as indicated in the description.

Limitations

• Row count, column definitions, and sample data are unknown, limiting suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

ONESTRUCTION on Hugging Face.

Collection Method

Likely created as a benchmark for automated IDS generation.

Freshness

Last updated 2026-05-18 11:33:36; freshness should be verified.