[论文] UI-Venus-2 Technical Report

论文概要

研究领域: NLP 作者: Venus Team, Zhuohan Cai, Haoxing Chen 发布时间: 2026-09-03 arXiv: 2509.00008

中文摘要

多模态GUI智能体已成为数字任务自动化的有前景范式，然而从面向基准的模型过渡到可靠的实际应用仍然具有挑战性，原因在于环境覆盖有限、任务构建脆弱以及奖励验证不可靠。在本工作中，我们提出了UI-Venus-2，一个通用基础GUI智能体，旨在通过统一的闭环推理-动作框架在移动、Web和桌面环境中运行。为了弥合走向实际部署的差距，我们联合扩展了三个关键维度：（1）环境，将覆盖范围扩展到170多个多语言移动应用和原生桌面操作系统；（2）任务，采用深度研究管道进行功能扎根的指令生成；（3）验证，采用具有视觉关键点和多模型投票的追踪级和样本级评估器，以确保可靠的RL训练信号。此外，我们集成了安全感知机制，以确保对重要动作的控制执行。通过提供一个能干、高效且开源的基础，UI-Venus-2推动了该领域向更通用、可验证和自反思的实际应用智能体发展。

原文摘要

Multimodal GUI agents have emerged as a promising paradigm for digital task automation, yet transitioning from benchmark-oriented models to dependable real-world applications remains challenging due to limited environment coverage, brittle task construction, and unreliable reward verification. In this work, we present UI-Venus-2, a general-purpose foundation GUI agent designed to operate across mobile, web, and desktop environments through a unified closed-loop reasoning-action framework. To bridge the gap toward practical deployment, we jointly scale three critical dimensions: (1) Environments, expanding coverage to more than 170 multilingual mobile apps and native desktop operating systems; (2) Tasks, employing a deep-research pipeline for function-grounded instruction generation; and (3...

#论文 #arXiv #NLP #小凯