PaulineLi/QuantiPhy-validation

Tasks

• video-text-to-text
• visual-question-answering

Citation text from the card

If you use this validation set in your work, please cite:

```bibtex
@article{li2025quantiphy,
  title   = {QuantiPhy: A Quantitative Benchmark Evaluating Physical Reasoning Abilities of Vision-Language Models},
  author  = {Li, Puyin and Xiang, Tiange and Mao, Ella and Wei, Shirley and Chen, Xinye and Masood, Adnan and Li, Fei-Fei and Adeli, Ehsan},
  journal = {arXiv preprint arXiv:2512.19526},
  year    = {2025}
}

## Measurement Conventions

- **Wingspan / width of flying animals**: the fully spread, wing-tip-to-wing-tip extent, i.e. the maximum width attained during flight.
- **Body length of animals** (e.g. the prior "length of the small bird"): the object's axis-aligned bounding-box length in the simulator.
- **`depth_info` for lab clips with an `s`/`x` suffix**: these are SAM-masked variants of the base clip; their depth references describe the original scene and may name objects that are masked out of the variant.

## Changelog

- 2026-10-07: fixed spelling errors in several questions; `video_source` value `segmentaion` corrected to `segmentation`; removed the BOM, trailing empty columns and stray `#NAME?` cells from `validation_dataset.csv`. Ground-truth values and row order are unchanged.
- 2026-10-08 (erratum): removed the row "What is the shoulder breadth of the person in cm?" listed under `captured_0027s` (index 2607). That clip is a SAM-masked variant in which the person is not visible, so the question could not be answered; it belonged to the base clip, which 

All datasets marked Permissive or share-alike | JSON