---
license: cc-by-4.0
language:
- en
tags:
- evaluation
- text-to-image
pretty_name: SpatialBench-UC
---

<img src="assets/banner.png" alt="SpatialBench-UC banner" width="100%"/>

<br/>

# SpatialBench-UC

<p align="center">
  <a href="https://arxiv.org/abs/2601.13462">
    <img src="https://img.shields.io/badge/arXiv-2601.13462-b31b1b.svg">
  </a>
  <a href="https://github.com/amineross/SpatialBench-UC">
    <img src="https://img.shields.io/badge/Code-GitHub-black.svg">
  </a>
</p>

<p align="center">
  Benchmark for <strong>pairwise spatial relations</strong> in text-to-image generation<br/>
  with <strong>explicit uncertainty</strong> and <strong>abstention</strong>
</p>

## Paper

**SpatialBench-UC: Uncertainty-Aware Evaluation of Spatial Prompt Following in Text-to-Image Generation**  
Amine Rostane  

https://arxiv.org/abs/2601.13462

## Code

https://github.com/amineross/SpatialBench-UC

## Citation

```bibtex
@misc{rostane2026spatialbenchucuncertaintyawareevaluationspatial,
      title={SpatialBench-UC: Uncertainty-Aware Evaluation of Spatial Prompt Following in Text-to-Image Generation}, 
      author={Amine Rostane},
      year={2026},
      eprint={2601.13462},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2601.13462}, 
}
```