Web-Shepherd: Advancing PRMs for Reinforcing Web Agents

Web-Shepherd is a process reward model for assessing individual steps of web-navigation trajectories. It introduces a collection of 40,000 step-level preference pairs with checklists and a separate benchmark for comparing reward models, then evaluates verification during WebArena-lite navigation.

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

web

Topics

reinforcement-learning, evaluation, data-generation, web-navigation

Record publication date

2025-05-21

Evidence retrieved

2026-10-16T06:36:54Z

Resource ID

arxiv:2505.15277

Recorded paper version

arXiv v2

Version updated: 2025-11-25T21:34:57Z.

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

Authors

Hyungjoo Chae, Sunghwan Kim, Junhee Cho, Seungone Kim, Seungjun Moon, Gyeom Hwangbo, Dongha Lim, Minjin Kim, Yeonjun Hwang, Minju Gwak, Dongwook Choi, Minseok Kang, Gwanhoon Im, ByeongUng Cho, Hyojun Kim, Jun Hee Han, Taeyoon Kwon, Minju Kim, Beong-woo Kwak, Dongjin Kang, Jinyoung Yeo

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

The annotation and scope were reviewed against the primary paper abstract. Experimental results and released artifacts have not been independently reproduced.

Connected assets

These links come from reviewed resource records. A documentation link does not imply a training-data or evaluation dependency.

• Documented resource: WebShepherd 3B · Models and checkpoints · relationship evidence
• Documented resource: WebPRM Collection preference data · Datasets · relationship evidence
• Documented resource: WebRewardBench · Benchmarks · relationship evidence

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 2026-10-16T06:36:54Z. Supports: name, authors, published_at, updated_at, identifiers.arxiv, details.arxiv_snapshot.

   Content fingerprint

   4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089
2. editorial primary review

   Retrieved 2026-10-16T06:36:54Z. Supports: name, kind, summary, environments, topics, scope, relations.

   Content fingerprint

   4953a8e0f029016aef668bce98cec4d2873a7ae49c1fe33c4311fa5eb8d24089

Suggest a correction

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