analiza aiqbench
This research introduces Teach VLM, a model that translates mobile screen trajectories into step-wise operational knowledge using demonstration videos, addressing the challenge of heterogeneous UI designs in vision-language models. By establishing a data flywheel for scalable acquisition and introducing the Teach-and-Repeat paradigm, the work provides an interpretable procedural reference to significantly improve task success rates for downstream GUI agents on Android devices.
Teach-and-Repeat: Accurately Extracting Operational Knowledge from Mobile Screen Demonstrations to Empower GUI Agents
arXiv:2606.12817v1 Announce Type: new Abstract: Understanding the digital world on mobile devices is shifting from static UI perception to dynamic action comprehension. This capability enables models to convert visual state transitions into operational knowledge, defined as short natural-language sentences that describe action types, target UI elements, textual arguments, and execution orders. However, due to the highly diverse and heterogeneous UI designs across applications, existing vision-language models (VLMs) struggle to accurately infer these underlying operations. To bridge this gap, we introduce Teach VLM, a core model designed to translate mobile screen trajectories into step-wise operational knowledge by extracting and analyzing operation-related keyframes from demonstration videos. To address the scarcity of aligned training data, we develop a systematic data flywheel for scalable data acquisition. We further introduce a novel Chinese Mobile Screen Teach Benchmark for fine-grained evaluation. Building upon Teach VLM, we propose the Teach-and-Repeat paradigm, where the generated operational knowledge serves as an interpretable procedural reference to guide downstream screen-based execution agents. Extensive evaluations demonstrate that Teach VLM significantly outperforms strong VLM baselines, achieving state-of-the-art performance in operation semantics prediction. Furthermore, experiments in Android World show that our paradigm yields consistent Task Success Rate improvements for downstream agents. Together, Teach VLM and the Teach-and-Repeat paradigm offer a practical pathway from raw demonstrations to reusable task automation.
Authors:Yudong Zhang (1), Lei Hu (1), Daoyang Liu (2), Jiawei Liu (1), Yangfan Luo (1), Xingyu Liu (1), Zuojian Wang (1), Zhilin Gao (1) ((1) Honor Device Co., Ltd, (2) The Chinese University of Hong Kong, Hong Kong, China)
Abstract:Understanding the digital world on mobile devices is shifting from static UI perception to dynamic action comprehension. This capability enables models to convert visual state transitions into operational knowledge, defined as short natural-language sentences that describe action types, target UI elements, textual arguments, and execution orders. However, due to the highly diverse and heterogeneous UI designs across applications, existing vision-language models (VLMs) struggle to accurately infer these underlying operations. To bridge this gap, we introduce Teach VLM, a core model designed to translate mobile screen trajectories into step-wise operational knowledge by extracting and analyzing operation-related keyframes from demonstration videos. To address the scarcity of aligned training data, we develop a systematic data flywheel for scalable data acquisition. We further introduce a novel Chinese Mobile Screen Teach Benchmark for fine-grained evaluation. Building upon Teach VLM, we propose the Teach-and-Repeat paradigm, where the generated operational knowledge serves as an interpretable procedural reference to guide downstream screen-based execution agents. Extensive evaluations demonstrate that Teach VLM significantly outperforms strong VLM baselines, achieving state-of-the-art performance in operation semantics prediction. Furthermore, experiments in Android World show that our paradigm yields consistent Task Success Rate improvements for downstream agents. Together, Teach VLM and the Teach-and-Repeat paradigm offer a practical pathway from raw demonstrations to reusable task automation.
20 pages, 9 figures. Yudong Zhang and Lei Hu contributed equally to this work. Xingyu Liu, Zuojian Wang, and Zhilin Gao are corresponding authors
Artificial Intelligence (cs.AI)
arXiv:2606.12817 [cs.AI]
(or arXiv:2606.12817v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2606.12817 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 291 cuvinte