Guardado en:
Detalles Bibliográficos
Autores principales: Yuan, Peng, Yin, Yuyang, Cai, Yuxuan, Wei, Zheng
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:https://arxiv.org/abs/2604.10988
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913025347289088
author Yuan, Peng
Yin, Yuyang
Cai, Yuxuan
Wei, Zheng
author_facet Yuan, Peng
Yin, Yuyang
Cai, Yuxuan
Wei, Zheng
contents Existing browser agent benchmarks face a fundamental trilemma: real-website benchmarks lack reproducibility due to content drift, controlled environments sacrifice realism by omitting real-web noise, and both require costly manual curation that limits scalability. We present WebForge, the first fully automated framework that resolves this trilemma through a four-agent pipeline -- Plan, Generate, Refine, and Validate -- that produces interactive, self-contained web environments end-to-end without human annotation. A seven-dimensional difficulty control framework structures task design along navigation depth, visual complexity, reasoning difficulty, and more, enabling systematic capability profiling beyond single aggregate scores. Using WebForge, we construct WebForge-Bench, a benchmark of 934 tasks spanning 7 domains and 3 difficulty levels. Multi-model experiments show that difficulty stratification effectively differentiates model capabilities, while cross-domain analysis exposes capability biases invisible to aggregate metrics. Together, these results confirm that multi-dimensional evaluation reveals distinct capability profiles that a single aggregate score cannot capture. Code and benchmark are publicly available at https://github.com/yuandaxia2001/WebForge.
format Preprint
id arxiv_https___arxiv_org_abs_2604_10988
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle WebForge: Breaking the Realism-Reproducibility-Scalability Trilemma in Browser Agent Benchmark
Yuan, Peng
Yin, Yuyang
Cai, Yuxuan
Wei, Zheng
Artificial Intelligence
Computer Vision and Pattern Recognition
Existing browser agent benchmarks face a fundamental trilemma: real-website benchmarks lack reproducibility due to content drift, controlled environments sacrifice realism by omitting real-web noise, and both require costly manual curation that limits scalability. We present WebForge, the first fully automated framework that resolves this trilemma through a four-agent pipeline -- Plan, Generate, Refine, and Validate -- that produces interactive, self-contained web environments end-to-end without human annotation. A seven-dimensional difficulty control framework structures task design along navigation depth, visual complexity, reasoning difficulty, and more, enabling systematic capability profiling beyond single aggregate scores. Using WebForge, we construct WebForge-Bench, a benchmark of 934 tasks spanning 7 domains and 3 difficulty levels. Multi-model experiments show that difficulty stratification effectively differentiates model capabilities, while cross-domain analysis exposes capability biases invisible to aggregate metrics. Together, these results confirm that multi-dimensional evaluation reveals distinct capability profiles that a single aggregate score cannot capture. Code and benchmark are publicly available at https://github.com/yuandaxia2001/WebForge.
title WebForge: Breaking the Realism-Reproducibility-Scalability Trilemma in Browser Agent Benchmark
topic Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2604.10988