# ScreenQA

About 86,000 human-written question-answer pairs over 35,000 Rico mobile screenshots, plus ScreenQA Short and ComplexQA variants, for benchmarking screen content understanding.

- Catalog URL: https://awesomecua.github.io/resources/github-google-research-datasets-screen-qa/
- Original source: https://github.com/google-research-datasets/screen_qa
- Kind: dataset
- Review: curated
- Scope: core
- Environments: android
- Topics: screen-understanding
- Evidence checked: 2026-10-16T06:36:01Z
- ID: github:google-research-datasets/screen_qa

## Limitations

Inclusion does not establish reproducibility or endorsement.

- Record JSON: https://awesomecua.github.io/resources/github-google-research-datasets-screen-qa/record.json

## Structured profile

- formats: JSON
- modalities: questions, answers, screen references, bounding boxes
- splits: train, validation, test
- supervision: screen question answering, answer grounding

### Count observations

- 68,951 question-answer pairs — original ScreenQA train split (reported; source: https://raw.githubusercontent.com/google-research-datasets/screen_qa/HEAD/README.md)
- 8,614 question-answer pairs — original ScreenQA validation split (reported; source: https://raw.githubusercontent.com/google-research-datasets/screen_qa/HEAD/README.md)
- 8,419 question-answer pairs — original ScreenQA test split (reported; source: https://raw.githubusercontent.com/google-research-datasets/screen_qa/HEAD/README.md)

All questions for a given screenshot share one split. ScreenQA Short reuses the same questions/screens; ComplexQA is a separate extension. Do not add these variants as independent original examples.

## License declarations

- code: not recorded
- data: not recorded
- document: not recorded
- weights: not recorded

## Primary evidence

- https://raw.githubusercontent.com/google-research-datasets/screen_qa/HEAD/README.md (retrieved 2026-10-16T06:36:01Z; SHA-256 7d1bfe7d6eddd5d3fc4c33d4814de3cabf7a6b6fd3d3f85ba4c2b33036bf8867)