Space Syntax-guided Post-training for Residential Floor Plan Generation
arXiv:2602.22507v1 Announce Type: new Abstract: Pre-trained generative models for residential floor plans are typically optimized to fit large-scale data distributions, which can under-emphasize critical architectural priors such as the configurational dominance and connectivity of domestic public spaces (e.g., living rooms and foyers). This paper proposes Space Syntax-guided Post-training (SSPT), a post-training paradigm that explicitly injects space syntax knowledge into floor plan generation via a non-differentiable oracle. The oracle converts RPLAN-style layouts into rectangle-space graphs through greedy maximal-rectangle decomposition and door-mediated adjacency construction, and then computes integration-based measurements to quantify public space dominance and functional hierarchy. To enable consistent evaluation and diagnosis, we further introduce SSPT-Bench (Eval-8), an out-of-distribution benchmark that post-trains models using conditions capped at $\leq 7$ rooms while eva
arXiv:2602.22507v1 Announce Type: new Abstract: Pre-trained generative models for residential floor plans are typically optimized to fit large-scale data distributions, which can under-emphasize critical architectural priors such as the configurational dominance and connectivity of domestic public spaces (e.g., living rooms and foyers). This paper proposes Space Syntax-guided Post-training (SSPT), a post-training paradigm that explicitly injects space syntax knowledge into floor plan generation via a non-differentiable oracle. The oracle converts RPLAN-style layouts into rectangle-space graphs through greedy maximal-rectangle decomposition and door-mediated adjacency construction, and then computes integration-based measurements to quantify public space dominance and functional hierarchy. To enable consistent evaluation and diagnosis, we further introduce SSPT-Bench (Eval-8), an out-of-distribution benchmark that post-trains models using conditions capped at $\leq 7$ rooms while evaluating on 8-room programs, together with a unified metric suite for dominance, stability, and profile alignment. SSPT is instantiated with two strategies: (i) iterative retraining via space-syntax filtering and diffusion fine-tuning, and (ii) reinforcement learning via PPO with space-syntax rewards. Experiments show that both strategies improve public-space dominance and restore clearer functional hierarchy compared to distribution-fitted baselines, while PPO achieves stronger gains with substantially higher compute efficiency and reduced variance. SSPT provides a scalable pathway for integrating architectural theory into data-driven plan generation and is compatible with other generative backbones given a post-hoc evaluation oracle.
Executive Summary
This article proposes Space Syntax-guided Post-training (SSPT), a novel approach to improving residential floor plan generation by integrating space syntax knowledge into pre-trained generative models. SSPT injects architectural priors such as configurational dominance and connectivity of domestic public spaces into floor plan generation, leading to improved public-space dominance and clearer functional hierarchy. The approach is scalable and compatible with other generative backbones, with two strategies: iterative retraining and reinforcement learning. Experiments demonstrate that SSPT outperforms distribution-fitted baselines, with PPO achieving stronger gains and higher compute efficiency. The SSPT-Bench evaluation framework provides a consistent evaluation and diagnosis protocol. This research has significant implications for the field of computer-aided design and architecture, offering a pathway for integrating theoretical knowledge into data-driven plan generation.
Key Points
- ▸ Space Syntax-guided Post-training (SSPT) integrates space syntax knowledge into pre-trained generative models for improved residential floor plan generation.
- ▸ SSPT injects architectural priors such as configurational dominance and connectivity of domestic public spaces into floor plan generation.
- ▸ Experiments demonstrate that SSPT outperforms distribution-fitted baselines, with PPO achieving stronger gains and higher compute efficiency.
Merits
Strength
SSPT provides a scalable pathway for integrating architectural theory into data-driven plan generation, making it a valuable contribution to the field of computer-aided design and architecture.
Interdisciplinary Approach
The integration of space syntax knowledge and generative models showcases an interdisciplinary approach to addressing complex design problems.
Evaluation Framework
The SSPT-Bench evaluation framework provides a consistent evaluation and diagnosis protocol, allowing for reproducibility and comparison of results.
Demerits
Limitation
The approach relies on pre-trained generative models, which may not capture the full complexity of architectural design problems.
Computational Requirements
The computational requirements of SSPT, particularly for reinforcement learning, may be substantial and require significant resources.
Expert Commentary
The article presents a novel and promising approach to integrating space syntax knowledge into pre-trained generative models for residential floor plan generation. The scalability and compatibility of SSPT with other generative backbones make it a valuable contribution to the field of computer-aided design and architecture. However, the approach relies on pre-trained generative models, which may not capture the full complexity of architectural design problems. Additionally, the computational requirements of SSPT may be substantial and require significant resources. Nevertheless, the SSPT-Bench evaluation framework provides a consistent evaluation and diagnosis protocol, allowing for reproducibility and comparison of results. Future research should focus on addressing the limitations and challenges of SSPT, as well as exploring its potential applications in other design fields.
Recommendations
- ✓ Further research should be conducted to explore the potential applications of SSPT in other design fields, such as urban planning and landscape architecture.
- ✓ The integration of SSPT with other generative models, such as GANs, should be investigated to address potential limitations and improve performance.
