# VisualWebArena

Visually grounded web tasks extending the WebArena evaluation setting.

- Catalog URL: https://awesomecua.github.io/resources/github-web-arena-x-visualwebarena/
- Original source: https://github.com/web-arena-x/visualwebarena
- Kind: benchmark
- Review: curated
- Scope: core
- Environments: web
- Topics: evaluation, screen-understanding
- Evidence checked: 2026-10-16T06:35:57Z
- ID: github:web-arena-x/visualwebarena

## Limitations

Inclusion does not establish reproducibility or endorsement.

- Record JSON: https://awesomecua.github.io/resources/github-web-arena-x-visualwebarena/record.json

## Structured profile

- actions: browser element actions
- evaluation: Execution-based completion checks for visually grounded web tasks.
- focus: multimodal web task completion
- observations: screenshots with set-of-mark annotations, caption-augmented accessibility tree
- setting: interactive

### Count observations

- 910 tasks — full VWA suite (reported; source: https://raw.githubusercontent.com/web-arena-x/visualwebarena/HEAD/README.md)
- 233 human trajectories — one task from each template type (reported; source: https://raw.githubusercontent.com/web-arena-x/visualwebarena/HEAD/README.md)

The 233 demonstration trajectories are not the full 910-task evaluation set. Observation modes and captioning components differ across baseline agents.

## License declarations

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

## Primary evidence

- https://raw.githubusercontent.com/web-arena-x/visualwebarena/HEAD/README.md (retrieved 2026-10-16T06:35:57Z; SHA-256 182529c2fce1e41f3a2162727205c54da8b3d2381b2273d726170ee9d8cf65cc)