AWS Case Study: Building a Domain-Specific BIM Model via 3-Stage Pipeline
AWS ML Blog · rss · 2026-08-12
The AWS ML Blog details how construction tech startup ONESTRUCTION built Ishigaki-IDS, a domain-specific foundation model for the BIM industry.
- Challenges: The IDS standard is new with scarce public data. The model needed to learn thousands of IFC vocabulary mappings and specific XML grammars, tasks where general LLMs score under 25% in structural compliance.
- 3-Stage Pipeline: Based on the Qwen open-source models, the team designed a CPT, SFT, and RLVR pipeline. They used experts to generate synthetic data and leveraged buildingSMART's audit tool as a reward function in RLVR to ensure mechanical correctness.
- Results: On their custom IDS-Bench, the domain model achieved nearly 100% in XML and IDS structural compliance and over 80% in content consistency, significantly outperforming general frontier models.
More from Research
- Research: Test-Time Compute Makes Overtraining Compute-Optimal — ajratner · 2026-08-12
- LLMs Are Breaking Science's 'Burden of Knowledge' Stagnation — emollick · 2026-08-12
- HALO: Human-AI collaborative system for drug discovery molecular hypothesis generation — _xiang_chen_ · 2026-08-12
- Solving RAG Bottlenecks: A 2026 Open-Source Guide to PDF Table Parsing — AvenueJay · 2026-08-12
- AI pushes mathematical bound for packing 17 squares to 4.456575 — stanislavfort · 2026-08-12
- Founder Uses ChatGPT to Design mRNA Cancer Vaccine for His Dog, Launches YC-Backed Gamgee — ycombinator · 2026-08-12