# Unified Vector Floorplan Generation via Markup Representation

Source: [Unified Vector Floorplan Generation via Markup Representation](<https://www.sota2.com/research/paper/69d48fc57d0172140ec74625>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-08T05:00:12.103Z

Authors: Kaede Shiohara, Toshihiko Yamasaki

Published: 2026-04-06T17:04:30.000Z

## Abstract

Automatic residential floorplan generation has long been a central challenge bridging architecture and computer graphics, aiming to make spatial design more efficient and accessible. While early methods based on constraint satisfaction or combinatorial optimization ensure feasibility, they lack diversity and flexibility. Recent generative models achieve promising results but struggle to generalize across heterogeneous conditional tasks, such as generation from site boundaries, room adjacency graphs, or partial layouts, due to their suboptimal representations. To address this gap, we introduce Floorplan Markup Language (FML), a general representation that encodes floorplan information within a single structured grammar, which casts the entire floorplan generation problem into a next token prediction task. Leveraging FML, we develop a transformer-based generative model, FMLM, capable of producing high-fidelity and functional floorplans under diverse conditions. Comprehensive experiments on the RPLAN dataset demonstrate that FMLM, despite being a single model, surpasses the previous task-specific state-of-the-art methods.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Conditional Floorplan Generation on RPLAN (test)](<https://www.sota2.com/research/sota/conditional-floorplan-generation-on-rplan-test.md>): Conditional Floorplan Generation; RPLAN (test); FID: 4.61.

- [Floorplan Generation on User Study Graph-conditional 100 samples (test)](<https://www.sota2.com/research/sota/floorplan-generation-on-user-study-graph-conditional-100-samples-test.md>): Floorplan Generation; User Study Graph-conditional 100 samples (test); Winning Rate: 51.

- [Floorplan Generation on RPLAN 5 rooms (test)](<https://www.sota2.com/research/sota/floorplan-generation-on-rplan-5-rooms-test.md>): Floorplan Generation; RPLAN 5 rooms (test); FID: 6.97.

- [Floorplan Generation on RPLAN 6 rooms (test)](<https://www.sota2.com/research/sota/floorplan-generation-on-rplan-6-rooms-test.md>): Floorplan Generation; RPLAN 6 rooms (test); FID: 5.2.

- [Floorplan Generation on RPLAN 7 rooms (test)](<https://www.sota2.com/research/sota/floorplan-generation-on-rplan-7-rooms-test.md>): Floorplan Generation; RPLAN 7 rooms (test); FID: 4.07.

- [Floorplan Generation on RPLAN 8 rooms (test)](<https://www.sota2.com/research/sota/floorplan-generation-on-rplan-8-rooms-test.md>): Floorplan Generation; RPLAN 8 rooms (test); FID: 4.64.

- [Floorplan Generation on RPLAN ALL (test)](<https://www.sota2.com/research/sota/floorplan-generation-on-rplan-all-test.md>): Floorplan Generation; RPLAN ALL (test); FID: 3.41.

- [Floorplan Generation on RPLAN (Efficiency Profile)](<https://www.sota2.com/research/sota/floorplan-generation-on-rplan-efficiency-profile.md>): Floorplan Generation; RPLAN; Training Time (h): 27.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.04859>)

[PDF](<https://arxiv.org/pdf/2604.04859.pdf>)