---
license: cc-by-4.0
---


This dataset supports the [From Vision To Language through Graph of Events in Space and Time: An Explainable Self-supervised Approach](https://arxiv.org/abs/2507.04815) paper.

### Dataset Description

The dataset comprises VLMs preference rankings of automatically generated video descriptions. For each video, the VLMs were asked to rank six descriptions based on criteria such as completeness and richness. 

## Citation 

```
@misc{masala2025visionlanguagegraphevents,
      title={From Vision To Language through Graph of Events in Space and Time: An Explainable Self-supervised Approach}, 
      author={Mihai Masala and Marius Leordeanu},
      year={2025},
      eprint={2507.04815},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2507.04815}, 
}
```