Guiding VLM Agents with Process Rewards at Inference Time for GUI Navigation

Guides visual GUI agents at inference time with a process-reward model that evaluates intermediate actions. It combines stepwise guidance with trajectory reflection and retry mechanisms in static and interactive navigation tasks.

AI-assisted editorial review against a primary source. This is a catalog annotation, not an independent reproduction or endorsement.

What is this resource?

Resource type

Research papers

Environment

unspecified

Topics

reward-models, self-reflection, grounding

Record publication date

2025-04-22

Evidence retrieved

2026-10-17T11:20:10Z

Resource ID

arxiv:2504.16073

Recorded paper version

arXiv v1

Version updated: 2025-04-22T17:52:42Z.

This identifies the retained arXiv metadata snapshot; newly observed changes require separate review.

Authors

Zhiyuan Hu, Shiyun Xiong, Yifan Zhang, See-Kiong Ng, Anh Tuan Luu, Bo An, Shuicheng Yan, Bryan Hooi

Cite this paper

Download BibTeX · Download CSL JSON

Citations preserve the selected source edition, its authors and recorded date precision. Unrecorded venues and identifiers are omitted.

Explore related topics

Scope and limitations

Reviewed against the exact-version official arXiv abstract page; its complete title, ordered author list, visible version, and normalized abstract SHA match the retained snapshot. The original API response body was not locally available. Experimental results and artifact availability have not been independently reproduced.

Primary evidence and verification

Source retrieval is separate from experiment reproduction. Use the original publication or publisher page when citing claims about the resource.

1. arxiv api

   Retrieved 2026-10-17T07:57:47Z. Supports: name, authors, published_at, updated_at, identifiers.arxiv, details.arxiv_snapshot.

   Content fingerprint

   27144160e8981bb9f36b60c391689774913ff3f026eeeffc2923154c404a6529
2. editorial primary review

   Retrieved 2026-10-17T11:20:10Z. Supports: name, kind, summary, environments, topics, scope, notes.

   Content fingerprint

   b1cf161fc20efc51363d9007952ce5f917963d4ab2ca32d39099229060b29528

Suggest a correction

Open a source-backed correction. Include this record ID and the original source. Read our editorial policy and coverage boundaries.