# VeriCUA: Learning Proactive Verification for Reliable GUI Agent Rewards

VeriCUA trains a Judge to select verification checks and stopping decisions while a frozen Probe inspects the post-task GUI. Single-attempt reinforcement learning balances verdict correctness and Probe use.

- Catalog URL: https://awesomecua.github.io/resources/arxiv-2610-26469/
- Original source: https://arxiv.org/abs/2610.26469
- Kind: paper
- Review: curated
- Scope: core
- Environments: desktop, linux, web, android
- Topics: reward-models, reinforcement-learning, evaluation, efficiency
- Evidence checked: 2026-10-20T08:27:26.236797+00:00
- ID: arxiv:2610.26469

## Limitations

Reviewed the complete abstract and selected v1 method/evaluation sections. The Judge is trained on OSWorld and frozen for MobileWorld/WebArena transfer; downstream Actor RL is separate. Evaluation uses screened terminal outcomes, and reported cost counts Probe tokens. No independent reproduction.

- Record JSON: https://awesomecua.github.io/resources/arxiv-2610-26469/record.json

## Authors

Zhizhou Wang, Yaxi Lu, Zhong Zhang, Zikun Hu, Xin Cong, Yankai Lin, Fangming Liu

## Citation exports

- BibTeX: https://awesomecua.github.io/resources/arxiv-2610-26469/citation.bib
- CSL JSON: https://awesomecua.github.io/resources/arxiv-2610-26469/citation.csl.json

## Recorded arXiv snapshot

- Version: 1
- Version source: https://arxiv.org/abs/2610.26469v1
- First submission: 2026-10-19T14:17:33Z
- Version updated: 2026-10-19T14:17:33Z
- Abstract SHA-256: 08ef7c9cdda652f60d3f2d4ab89e7218ad5a2b85f5c76f0277c289db06fd2796
- Abstract hash normalization: sha256-utf8-whitespace-collapsed-v1

## License declarations

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

## Primary evidence

- https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22browser+agent%22%29+AND+submittedDate%3A%5B202610060000+TO+202610200000%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending (retrieved 2026-10-20T05:52:00Z; SHA-256 af310824eea97aac8209023ff53bad7bea77778816586b10bb94110148347af9)
- https://export.arxiv.org/api/query?search_query=%28ti%3A%22computer+use%22+OR+ti%3A%22computer-use%22+OR+ti%3A%22GUI+agent%22+OR+ti%3A%22GUI+agents%22+OR+ti%3A%22GUI+grounding%22+OR+ti%3A%22web+agent%22+OR+ti%3A%22web+agents%22+OR+ti%3A%22browser+agent%22+OR+ti%3A%22browser+agents%22+OR+ti%3A%22mobile+agent%22+OR+ti%3A%22mobile+agents%22+OR+ti%3A%22desktop+agent%22%29+AND+submittedDate%3A%5B202610130000+TO+202610192359%5D&start=0&max_results=200&sortBy=submittedDate&sortOrder=descending (retrieved 2026-10-20T08:27:26.236797+00:00; SHA-256 80bed80786f939973d0d13b4f892300fee02a6110e326247131fc83d257796de)
- https://arxiv.org/html/2610.26469v1 (retrieved 2026-10-20T08:28:59.073204+00:00; SHA-256 f13f0225c721d5aa430bb8aea1d1e7a7da5973760c4c3b82f26ba648aade82ba)