---
license: other
license_name: flux-non-conmercial-license
license_link: LICENSE.md
language:
- en
tags:
- image-generation
- image-editing
- flux
- diffusion-single-file
pipeline_tag: image-text-to-image
library_name: diffusers
base_model:
- black-forest-labs/FLUX.2-klein-9B
datasets:
- YuhongLiu/SpatialASM-edit140k
---


# SpatialASM: Internalizing Active Mentalization for Spatial Reasoning</h1>

  📖<a href="http://arxiv.org/abs/2610.13832">Paper</a>
  | 🏠<a href="https://github.com/InternLM/SpatialASM">Homepage</a >
  | 🤗<a href="https://huggingface.co/collections/YuhongLiu/spatialasm">SpatialASM Collection</a >
  | 🤗<a href="https://huggingface.co/datasets/YuhongLiu/SpatialASM-128k">SpatialASM-128k Dataset</a >
  | 🤗<a href="https://huggingface.co/YuhongLiu/SpatialASM-NVG">SpatialASM-NVG Model</a >


## 📢 News
- 🚀 [2026/10/12] We have released the SpatialASM [📖Arxiv paper](https://arxiv.org/abs/2610.13832) and its [repository](https://github.com/InternLM/SpatialASM).
- 🚀 [2025/10/12] We have released our allocentric novel-view generator [🤗SpatialASM-NVG](https://huggingface.co/YuhongLiu/SpatialASM-NVG).
- 🚀 [2025/10/12] We have released [🤗SpatialASM-edit140k](https://huggingface.co/datasets/YuhongLiu/SpatialASM-edit140k), an image-text-to-image dataset for tuning novel-view generators, and two spatial-VQA datasets with high-quality CoT responses: [🤗SpatialASM-128k](https://huggingface.co/datasets/YuhongLiu/SpatialASM-128k) and its subset [🤗SpatialASM-66k](https://huggingface.co/datasets/YuhongLiu/SpatialASM-66k).

## 🌈 Overview

Humans rarely solve difficult spatial problems by passively inspecting the views in front of them. Instead, we revisit a scene from task-relevant viewpoints and mentally reconstruct what should be visible there. We call this capability **Active Spatial Mentalization (ASM)**.

We are thrilled to introduce **SpatialASM**, a scalable framework that internalizes ASM into LVLMs. Through the two-stage training on LVLMs, SpatialASM elicits active spatial exploration and imagination and enhances the overall spatial reasoning performance.
<p style="text-align: center;"> 
  <img src="Assets/teaser.png" alt="Teaser" width="100%"> 
</p>

SpatialASM consists of three complementary components:

1. **Allocentric Novel-View Generator.** SpatialASM-NVG renders the scene from a target viewpoint specified by an anchor image, a pixel location, and a viewing direction. This allocentric formulation avoids lengthy sequences of camera-centric movements.
2. **ASM Trajectory Synthesis.** Given a spatial QA sample, an LVLM plans informative viewpoints and the NVG renders the corresponding observations. After GeoP and answer-consistency filtering, rendered views are replaced by textual captions to form explicit plan-imagine-reason trajectories for SFT.
3. **View-Imagination Reinforcement Learning.** Geometry-filtered candidate views are used to construct automatically verifiable perspective-taking tasks. GRPO further strengthens the association between a viewpoint description and its expected observation.
<p style="text-align: center;"> 
  <img src="Assets/pipeline.png" alt="Pipeline" width="100%"> 
</p>


## 💡 Highlights

- 🔥 **Active spatial reasoning:** SpatialASM jointly internalizes task-conditioned **view planning** and viewpoint-conditioned **view imagination**, enabling LVLMs to spontaneously seek and mentally construct missing spatial evidence.
- 🔥 **Scalable spatial CoT supervision:** SpatialASM converts existing answer-only spatial QA data into explicit plan-imagine-reason trajectories and automatically verifiable perspective-taking tasks, alleviating the scarcity of spatial Chain-of-Thought supervision.
- 🔥 **Training-time spatial simulation:** A novel-view generator acts as temporary training scaffolding. The final model does not depend on a world model, renderer, or additional visual observations at inference time.
- 🔥 **Grounded and reliable supervision:** An allocentric viewpoint interface reaches a target perspective in one query, while the **Geometry Perturbation (GeoP)** metric and answer-consistency checks filter unreliable synthesized views and trajectories.
<p style="text-align: center;"> 
  <img src="Assets/comparison.png" alt="Comparison" width="100%"> 
</p>




## Cases
<p style="text-align: center;"> 
  <img src="Assets/NVG.png" alt="sample" width="90%"> 
</p>


## ✒️Citation
If you find this project useful, please kindly cite:
```
@article{liu2026spatialasm,
  title={SpatialASM: Internalizing Active Mentalization for Spatial Reasoning},
  author={Yuhong Liu, Shenxi Wu, Junyi Chi, Beichen Zhang, Xuanlang Dai, Jiaqi Wang, Yuhang Zang, Dahua Lin},
  journal={arXiv preprint arXiv:2610.13832},
  year={2026}
}
```

## 📄 License
This model falls under the [FLUX Non-Commercial License](https://huggingface.co/black-forest-labs/FLUX.2-klein-9B/blob/main/LICENSE.md).