Floora improves architectural layout design with specialized AI model
FLOORA: A Human-Aligned Domain-Specific Language Model for Architectural Design
Machine LearningArtificial Intelligence
Summary
Architectural design often needs careful arrangement of rooms and spaces, which general AI models struggle with. The authors present FLOORA, a smaller AI model tailored specifically for creating floor plans that follow architectural rules better. FLOORA learns from specialized data and human preferences to produce more valid and practical building layouts than larger models. Tests by both AI judges and human experts show it consistently makes better floor designs, even on new and unusual buildings. This approach may help create AI tools for other engineering fields that need precise, structured outputs.
What this means in practice
- •For architectural designers: Generate building floor plans that better conform to architectural constraints and human preferences using a specialized AI model.
- •For software engineers in engineering domains: Develop domain-specific AI models for other engineering areas that require structured and verifiable outputs similar to architectural layouts.
Authors
Sahand Rezaei-Shoshtari, Patryk Wozniczka, Shu Ishida, Gregg Streuber, Farnoosh Javadi, Jeffrey Landes, Angela Ju, Muhammad Azam, Bryan Lim, Johan Luttun, Indrajeet Haldar, Jonathan Shaw, Beatriz Guerra, Ivan Sosnovik, James Stoddart, Robert Giaquinto, Adam Gaier
Abstract
Foundation models are powerful generators, but many engineering domains require structured representations that general-purpose systems handle poorly. We introduce FLOORA (Floor Layout Optimization with RL Alignment), a family of small domain-specific language (DSL) models for architectural layout generation. With specialized data and alignment, our 0.6B model outperforms much larger frontier models, achieving VLM judge win rates up to 92.0% on out-of-distribution real-world buildings and 96.0% on synthetic buildings. Human evaluations further corroborate these results, with FLOORA selected as the best model in 89.3% of evaluations. FLOORA combines a token-efficient DSL, custom tokenization, domain-specific pretraining, supervised fine-tuning (SFT), and reinforcement learning (RL) with learned human-preference and verifiable rewards. This pipeline improves architectural and geometric validity, supported by extensive empirical evaluation and ablation studies. Although focused on architecture, our results suggest that similar domain-specific recipes may be useful in other engineering domains with structured, verifiable outputs. Datasets, models, and inference code are available at https://github.com/AutodeskAILab/floora.