README.md

---
license: apache-2.0
task_categories:
- question-answering
language:
- en
arxiv_id: 2509.25390
pretty_name: SpinBench
---

<div align="center">
  <h1><img src="assets/spinbench_logo.png" width="50" /> SpinBench: Perspective and Rotation as a Lens on Spatial Reasoning in VLMs </h1>
</div>

<h5 align="center">
  <a href="https://spinbench25.github.io/">🌐 Project page</a> |
  <a href="https://huggingface.co/datasets/YuyouZhang/SpinBench">🤗 Dataset</a> |
  <a href="https://arxiv.org/abs/2509.25390">📑 Paper</a> |
  <a href="https://github.com/ZhangYuyou-10/SpinBench">💻 Code</a>
</h5>

<!-- [Project page](https://spinbench25.github.io/) • [arXiv:2509.25390](https://arxiv.org/abs/2509.25390) -->

**SpinBench** is a cognitively grounded diagnostic benchmark for evaluating **spatial reasoning** in vision-language models (VLMs).
![Alt text](assets/overview_example.png)
SpinBench is designed around the core challenge of spatial reasoning: perspective taking, the ability to reason about how scenes and object relations change under viewpoint transformation. Since perspective taking requires multiple cognitive capabilities, such as recognizing objects across views, relative positions grounding, and mentally simulating transformations, SpinBench introduces a set of fine-grained diagnostic categories. Our categories target translation, rotation, object relative pose, and viewpoint change, and are progressively structured so that single-object simpler tasks scaffold toward the most demanding multi-object perspective-taking setting. We evaluate 37 state-of-the-art VLMs, both proprietary and open source. Results reveal systematic weaknesses: strong egocentric bias, poor rotational understanding, and inconsistencies under symmetrical and syntactic reformulations. Scaling analysis shows both smooth improvements and emergent capabilities. While human subjects achieve high accuracy (91.2%), task difficulty as measured by human response time shows strong correlation with VLM accuracy, indicating that SpinBench captures spatial reasoning challenges shared across humans and VLMs. We believe SpinBench provides critical insights into spatial reasoning in VLMs and highlights key gaps in their ability to reason about physical space. 

---

## 📁 Dataset

<details>
<summary><strong>Click to expand folder structure</strong></summary>

&nbsp;

```
SpinBench/
├── test.jsonl
├── test_small.jsonl
├── images/
│   ├── cars_rotation_c187650a7b.jpg
│   ├── face_rotation_2b4fd309cf.png
│   ├── infinigen_d3f202e7a1.png
│   ├── original_01bce239aa.jpg
│   └── ...
├── README.md
└── ...
```

</details>

## Citation

**BibTeX:**
```bibtex
@article{zhang2025spinbench,
  title={SpinBench: Perspective and Rotation as a Lens on Spatial Reasoning in VLMs},
  author={Zhang, Yuyou and Corcodel, Radu and Hori, Chiori and Cherian, Anoop and Zhao, Ding},
  journal={arXiv preprint arXiv:2509.25390},
  year={2025}
}
```