Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

From Pixels to Policies: Reinforcing Spatial Reasoning in Language Models for Content-Aware Layout Design

Sha Li, Stefano Petrangeli, Yu Shen, Xiang Chen

TL;DR

LaySPA is introduced, a reinforcement learning framework that equips large language models (LLMs) with explicit and interpretable spatial reasoning for content-aware graphic layout design and improves structural validity and visual quality.

Abstract

We introduce LaySPA, a reinforcement learning framework that equips large language models (LLMs) with explicit and interpretable spatial reasoning for content-aware graphic layout design. LaySPA addresses two key challenges: LLMs' limited spatial reasoning and the lack of opacity in design decision making. Instead of operating at the pixel level, we reformulate layout design as a policy learning problem over a structured textual spatial environment that explicitly encodes canvas geometry, element attributes, and inter-element relationships. LaySPA produces dual-level outputs comprising interpretable reasoning traces and structured layout specifications, enabling transparent and controllable design decision making. Layout design policy is optimized via a multi-objective spatial critique that decomposes layout quality into geometric validity, relational coherence, and aesthetic consistency, and is trained using relative group optimization to stabilize learning in open-ended design spaces. Experiments demonstrate that LaySPA improves structural validity and visual quality, outperforming larger proprietary LLMs and achieving performance comparable to specialized SOTA layout generators while requiring fewer annotated samples and reduced latency.

From Pixels to Policies: Reinforcing Spatial Reasoning in Language Models for Content-Aware Layout Design

TL;DR

LaySPA is introduced, a reinforcement learning framework that equips large language models (LLMs) with explicit and interpretable spatial reasoning for content-aware graphic layout design and improves structural validity and visual quality.

Abstract

We introduce LaySPA, a reinforcement learning framework that equips large language models (LLMs) with explicit and interpretable spatial reasoning for content-aware graphic layout design. LaySPA addresses two key challenges: LLMs' limited spatial reasoning and the lack of opacity in design decision making. Instead of operating at the pixel level, we reformulate layout design as a policy learning problem over a structured textual spatial environment that explicitly encodes canvas geometry, element attributes, and inter-element relationships. LaySPA produces dual-level outputs comprising interpretable reasoning traces and structured layout specifications, enabling transparent and controllable design decision making. Layout design policy is optimized via a multi-objective spatial critique that decomposes layout quality into geometric validity, relational coherence, and aesthetic consistency, and is trained using relative group optimization to stabilize learning in open-ended design spaces. Experiments demonstrate that LaySPA improves structural validity and visual quality, outperforming larger proprietary LLMs and achieving performance comparable to specialized SOTA layout generators while requiring fewer annotated samples and reduced latency.
Paper Structure (31 sections, 10 equations, 6 figures, 2 tables)

This paper contains 31 sections, 10 equations, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Problem Formulation
  4. LaySPA
  5. Spatially-Grounded Environment
  6. Dual-Level Design Artifacts
  7. Multi-Objective Spatial Critique
  8. 1) Format Correctness Reward $R_{\text{format}}$
  9. 2) Layout Quality Reward $R_{\text{quality}}$
  10. 3) IoU Matching Reward $R_{\text{IoU}}$
  11. 4) Final Hybrid Reward
  12. Training Procedure
  13. Experiments
  14. Datasets
  15. Experimental Setup
  16. ...and 16 more sections

Figures (6)

  • Figure 1: Comparison between a human-designed poster and a GPT-5–generated layout, illustrating common spatial reasoning failures in LLM-based layout generation.
  • Figure 2: The framework and workflow of LaySPA.
  • Figure 3: Comparison of Qwen-7B w/wo LaySPA on CGL and PKU. Red highlights improvements.
  • Figure 4: Layouts generated by different methods.
  • Figure 5: Ablation study on varying reward weights
  • ...and 1 more figures