Spatial Chain-of-Thought: Bridging Understanding and Generation Models for Spatial Reasoning Generation

Wei Chen1 Yancheng Long2 Mingqiao Liu3 Haojie Ding4 Yankai Yang2 Hongyang Wei3 Yi-Fan Zhang5 Bin Wen4 Fan Yang4 Tingting Gao4 Han Li4 Long Chen1
1HKUST   2HIT SZ   3THU   4Kuaishou   5CASIA
arXiv Code (Soon) Model (Soon) Dataset (Soon)
Teaser image (Figure 1)

Figure 1. Comparison of conditioning interfaces for spatially constrained text-to-image generation. (a) Continuous Bridge: MLLM encodes the prompt into continuous features and feeds them to the generation model. (b) Textual Bridge: MLLM expands the prompt into textual CoT, but spatial layout is compressed into language, and fine-grained structure is easily lost. (c) Spatial CoT Bridge (ours): MLLM outputs a Spatial CoT with object-level bounding boxes, which are rendered to an image via interleaved text–coordinates into a spatially-aware diffusion model, enabling faithful synthesis under strict spatial rules (e.g., a 3×4 desk grid with required empty neighbors).

Abstract

While diffusion models have shown exceptional capabilities in aesthetic image synthesis, they often struggle with complex spatial understanding and reasoning. Existing approaches resort to Multimodal Large Language Models (MLLMs) to enhance this capability. However, they either incur high computational costs through joint training or suffer from spatial information loss when relying solely on textual prompts. To alleviate these limitations, we propose a Spatial Chain-of-Thought (SCoT) framework, a plug-and-play approach that effectively bridges the reasoning capabilities of MLLMs with the generative power of diffusion models. Specifically, we first enhance the diffusion model's layout awareness by training it on an interleaved text-coordinate instruction format. We then leverage state-of-the-art MLLMs as planners to generate comprehensive layout plans, transferring their spatial planning capabilities directly to the generation process. Extensive experiments demonstrate that our method achieves state-of-the-art performance on image generation benchmarks and significantly outperforms baselines on complex reasoning tasks, while also showing strong efficacy in image editing scenarios.

Method

Spatial CoT Bridge. An off-the-shelf MLLM planner converts a user prompt into an interleaved text–bounding-box Spatial Chain-of-Thought plan, which is then rendered by a spatially-aware diffusion model.

Pipeline (Figure 2)

Figure 2. Overview of our plug-and-play framework (MLLM-based planner + spatially-aware renderer).

Key Contributions

  • We propose Spatial Chain-of-Thought, a spatial planning process that bridges MLLM planning and diffusion rendering in an efficient and plug-and-play way.
  • We build a spatially-aware generation model trained with interleaved text–coordinate instructions, and pair it with an MLLM-based planner that outputs executable bbox-specified layouts for complex spatial constraints.
  • We conduct extensive experiments on text-to-image generation and editing benchmarks, showing that our method significantly outperforms baselines in complex scenarios.

Results

Qualitative comparisons.

Qualitative comparison

Figure 3. Qualitative comparison in complex spatial scenes across SOTA models and ours.

BibTeX

@misc{chen2026spatialchainofthoughtbridgingunderstanding,
      title={Spatial Chain-of-Thought: Bridging Understanding and Generation Models for Spatial Reasoning Generation}, 
      author={Wei Chen and Yancheng Long and Mingqiao Liu and Haojie Ding and Yankai Yang and Hongyang Wei and Yi-Fan Zhang and Bin Wen and Fan Yang and Tingting Gao and Han Li and Long Chen},
      year={2026},
      eprint={2602.11980},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2602.11980}, 
}