react native development

本文通过对比五个植物管理应用实现（iOS、Android原生及Flutter等跨平台框架），评估了软件质量的权衡，使用ISO/IEC 25010标准衡量不同维度性能。

Existing mobile agents are constrained by the graphical user interface (GUI) paradigm, struggling to efficiently handle complex tasks such as batch operations and cross-application workflows. This work presents the first systematic exploration of command-line interfaces (CLIs) as an alternative interaction paradigm, enabling direct invocation of device services and data without requiring screen perception or touch-based actions. We introduce CLI-Advantage, a benchmark suite encompassing five representative scenarios challenging for GUI-based approaches, along with open-sourced evaluation infrastructure. Experimental results demonstrate that, without any mobile-specific fine-tuning, general-purpose code large language models—such as Claude Code—combined with Android CLI toolchains achieve success rates of 71.8% on AndroidWorld and 51.9% on MobileWorld, substantially outperforming current GUI-based baselines. Moreover, these CLI-driven agents accomplish tasks in an average of 10.7 steps, significantly fewer than the 18.6 steps required by GUI counterparts.

该研究通过引入MobileForge基准，解决了多屏移动应用生成中的跨页导航、代码可维护性等问题，评估了现有模型在项目级生成中的表现。

To address semantic misjudgment in cross-app Android GUI testing caused by operational logic discrepancies—and the poor generalizability of existing methods (e.g., mapping-based or LLM-based approaches) that rely on logical consistency—this paper proposes the first multi-agent collaborative, vision-language model (VLM)-driven framework. Our method jointly processes visual and textual modalities of GUIs, performs stage-wise semantic parsing of UI elements, emphasizes functional commonality rather than exact path matching, and incorporates an active feedback mechanism to enhance transfer robustness. Evaluated on the LinPro dataset (39 apps, 578 tasks), our framework achieves a 90.3% cross-app transfer success rate—outperforming the best mapping-based and LLM-based baselines by 318.1% and 109.1%, respectively. This marks a significant breakthrough in overcoming transfer bottlenecks under operational logic heterogeneity.

GUI test transfer across applications faces challenges such as mapping failure and incomplete test coverage due to structural heterogeneity among UI controls. This paper proposes a novel “abstraction–concretization” paradigm: instead of performing interface-level control mapping, it first abstracts universal test logic from multi-source, functionally equivalent test cases; then, leveraging a large language model (LLM), it concretizes this logic into executable test scripts for the target app via functional semantic alignment and joint GUI event–assertion generation. To our knowledge, this is the first approach to deeply integrate LLMs throughout both test logic generalization and target-specific adaptation, enabling functional-semantic-level transfer. Evaluated on the FrUITeR and Lin datasets, our method achieves functional test success rates of 64% (+191%) and 75% (+42%), respectively—demonstrating substantial improvements in transfer effectiveness and practicality.

This study addresses the limited cross-application generalization of mobile GUI agents and the inadequacy of existing benchmarks in evaluating this deficiency. To this end, it proposes AnyAppBench, a real-time Android benchmark spanning 52 applications that assesses agent performance across heterogeneous interfaces through category control and fixed-target testing. By integrating VLM-as-a-judge evaluation, automated task template generation, and human-in-the-loop annotation, the work establishes a systematic failure taxonomy. The contributions include quantitatively revealing cross-application generalization gaps among thirteen agents, demonstrating that success on source applications is difficult to transfer, and showing that subgoal decomposition strategies yield limited effectiveness. These findings provide critical analytical foundations for improving the robust deployment of mobile GUI agents.

This study addresses the challenges of user attrition and limited experimental flexibility in native application rewrites by proposing a Strangler Fig pattern based on a dual-launch mechanism for native platforms. This approach hosts multiple version variants within a single binary, enabling dynamic runtime selection and full lifecycle management through symbol resolution mapping and linker retention lists. As the first application of this pattern to native mobile environments, it facilitates binary-level A/B testing and seamless legacy deprecation. Empirical results from an iOS application rewrite demonstrate that the migration was completed within ten months with zero user churn, significantly enhancing both development efficiency and experimentation capabilities.

研究了跨生态系统软件包的普遍性、架构模式及其与项目健康度的关系，通过分析六大生态系统中的六百万个软件包，识别出五种架构模式。

This work addresses the challenge of repairing visual defects in Android applications, which is often hindered by incomplete human-submitted bug reports lacking observed behavior (OB), expected behavior (EB), or steps to reproduce (S2Rs). To overcome this limitation, the authors propose the first approach that leverages GUI context to guide large language models (LLMs) in generating structured bug reports. By integrating interaction logs and screenshots, the method automatically produces comprehensive and accurate OB, EB, and S2Rs. A unified evaluation framework assessing both correctness and completeness is also introduced. Experimental results on 48 bug reports across 26 applications demonstrate that the generated reports significantly outperform both original human-written reports and existing LLM-based baselines in quality.

This study addresses the navigation inefficiency of existing mobile GUI agents that rely on screen-by-screen interactions. To handle complex tasks, this work proposes GUI-Hopper, a hybrid interaction architecture that integrates deeplink-based direct navigation with conventional GUI operations. The core innovation lies in a novel method for constructing a deeplink catalog through static code analysis and automated on-device verification, combined with large language model planning to enable hybrid action space modeling. Experimental results demonstrate that this approach significantly improves both task success rates and execution efficiency in real-world commercial applications, validating the effectiveness of the proposed hybrid interaction paradigm.