SpatialTunnel

SpatialTunnel is a Blender-rendered diagnostic dataset for studying how vision-language models represent spatial relations internally. It was introduced in Why Far Looks Up: Probing Spatial Representation in Vision-Language Models (arXiv:2605.30161).

Resources

• Project page
• Contrastive-probing code
• SpatialTunnel generation code

Dataset Configurations

Config	File	Rows	Description
phase_variation	phase_variation-*.parquet	12,288	Binary depth-comparison questions with controlled angular-position variation.
size_variation	size_variation-*.parquet	4,400	Binary depth-comparison questions under controlled object-size variation.
contrastive_probing	contrastive_probing.parquet	1,200	Balanced spatial-relation questions for contrastive probing.

Format

All configurations use Parquet files with the following columns:

For phase_variation and size_variation, each rendered image is paired with four binary VQA questions that vary the object order and comparison direction (closer/farther).

Column	Description
index	Row index within the selected config.
image	PNG image stored as a Hugging Face Image feature.
question	Spatial question to ask the model.
answer	Ground-truth answer for the row.

contrastive_probing.tsv is a TSV file for compatibility with the contrastive-probing code.

---

Citation

If you use this dataset, please cite our paper.

@article{min2026whyfarlooksup,
  title   = {Why Far Looks Up: Probing Spatial Representation in Vision-Language Models},
  author  = {Min, Cheolhong and Jung, Jaeyun and Lee, Daeun and Jeon, Hyeonseong and
             Su, Yu and Tremblay, Jonathan and Song, Chan Hee and Park, Jaesik},
  journal = {arXiv preprint arXiv:2605.30161},
  year    = {2026},
}