---
dataset_info:
  features:
  - name: question_no_arrow
    dtype: string
  - name: question_with_arrow
    dtype: string
  - name: question
    dtype: string
  - name: answer
    dtype: string
  - name: answer_choices
    dtype: string
  - name: question_type
    dtype: string
  - name: marked_image_no_arrow
    dtype: image
  - name: marked_image_with_arrow
    dtype: image
  - name: marked_image
    dtype: image
  - name: new_perspective
    dtype: image
  - name: scene_name
    dtype: string
  - name: trajectory_id
    dtype: string
  splits:
  - name: distance_change_closer
    num_bytes: 18849012438.825
    num_examples: 5525
  - name: distance_change_further
    num_bytes: 6463104122.9
    num_examples: 1850
  - name: relative_position_left_left
    num_bytes: 4537034737.872
    num_examples: 1288
  - name: relative_position_left_right
    num_bytes: 18537627601.4
    num_examples: 5200
  - name: relative_position_right_left
    num_bytes: 18882980157.978
    num_examples: 5406
  - name: relative_position_right_right
    num_bytes: 4498574922.318
    num_examples: 1262
  download_size: 70968763871
  dataset_size: 71768333981.293
configs:
- config_name: default
  data_files:
  - split: distance_change_closer
    path: data/distance_change_closer-*
  - split: distance_change_further
    path: data/distance_change_further-*
  - split: relative_position_left_left
    path: data/relative_position_left_left-*
  - split: relative_position_left_right
    path: data/relative_position_left_right-*
  - split: relative_position_right_left
    path: data/relative_position_right_left-*
  - split: relative_position_right_right
    path: data/relative_position_right_right-*
---


## Citation

Released with the paper **Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models** ([arXiv:2606.03988](https://arxiv.org/abs/2606.03988)):

```bibtex
@misc{bigverdi2026imaginativeperceptiontokensenhance,
      title={Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models},
      author={Mahtab Bigverdi and Linjie Li and Weikai Huang and Yiming Liu and Jaemin Cho and Jieyu Zhang and Tuhin Kundu and Chris Dangjoo Kim and Zelun Luo and Linda Shapiro and Ranjay Krishna},
      year={2026},
      eprint={2606.03988},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2606.03988},
}
```