UI-Venus-2, an AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → for mobile, web and desktop interfaces, checks recorded interactions against visible evidence when selecting training data. Its technical report describes breaking a task into observable milestones and examining screenshots to determine which were reached.
For example, a trace that enters the right group and fills in a post but runs out of steps before confirming publication is labeled partial rather than failed. Such traces can be reviewed and repaired, while completed traces enter a higher-quality candidate pool. The mechanism separates unfinished but useful interactions from successful ones before they are used to develop the agent.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
articleWebUIProof: Benchmarking WebUI Code Generators with UI-Agent Execution HarnessYun-Yun Tsai, Yuning Mao, Shiqi Wang, Junfeng Yang, Sinong Wang
articleGUI-CC: Benchmarking Contextual Consistency of GUI World Models as Agent EnvironmentsLin Fu, Zheyuan Yang, Tianhui Zhang, Jinbiao Wei, Guo Gan, Boxu Liu, Yilun Zhao, Yu Rong
articleOpenMobile-2: Building Versatile Mobile Agents with Scalable Environments and App-Native ToolsChenyang Yan, Yingying Zhang, Kanzhi Cheng, Qiushi Sun, Zhengyuan Pan, Zheng Ma, Hang Yan, Yian Wang, Nuo Chen, Jialin Cao, Xingdong Gong, Wenpo Song, Han Chen, Zichen Ding, Fangzhi Xu, Shujian Huang, Xinyu Dai, Yichen Liu, Tiankuo Yao, Bo Wang, Ben Kao, Jianbing Zhang, Lewei Lu, Dahua Lin