Newest papers

LiveUIBench: Evaluating Agent as App Workflows

Tong Zhang, Yuxuan Cheng, Ruimin Zhang, Yun Peng, Tao Xie · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.27977 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

Aplikasi menyediakan fitur yang nyaman bagi pengguna untuk menyelesaikan tugas sehari-hari. Namun, pengguna yang berbeda mungkin memiliki persyaratan yang berbeda atau bahkan bertentangan untuk aplikasi yang sama, sementara persyaratan pengguna yang sama dapat berubah dari waktu ke waktu. Personalisasi aplikasi yang diproduksi oleh pendekatan agen-to-app atau UI-to-code dapat memerlukan perubahan pada implementasi yang dihasilkan. Kami mengusulkan Agen sebagai App, tugas di mana pengguna mengekspresikan kebutuhan mereka dalam bahasa alami dan agen membuat aplikasi yang dipersonalisasi sambil terus memberikan perilaku selama penggunaan. Kami menerapkan LiveUIAgent, yang menghasilkan orkestrasi spesifik tugas yang menghubungkan antarmuka pengguna, informasi yang disimpan, dan tindakan layanan. Agen bersama memodifikasi runtime dan memperbarui antarmuka saat melakukan operasi yang diminta. Meninjau orkestrasi dan antarmuka sambil menggunakan kembali runtime yang memberikan dasar

Machine translation (Meta M2M100); the English original follows.

Abstract

Applications provide convenient features for users to complete daily tasks. However, different users may have different or even conflicting requirements for the same application, while the requirements of the same user may change over time. Personalizing applications produced by agent-to-app or UI-to-code approaches can require changes to the generated implementation. We propose Agent as App, a task in which users express their requirements in natural language and an agent creates a personalized application while continuing to supply its behavior during use. We implement LiveUIAgent, which generates a task-specific orchestration connecting the user interface, saved information, and service actions. A shared agent runtime renders and updates the interface while carrying out the requested operations. Revising the orchestration and interface while reusing that runtime provides a lightweight basis for personalization as requirements evolve. We introduce LiveUIBench, a benchmark with diverse requirements and ordered interactions across personal services. The empirical findings show that the evaluated models struggle to complete these workflows reliably. Applications can satisfy many local checks while leaving the overall task unfinished, and greater partial coverage can coincide with fewer completed workflows.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Orchestration, Workflow, Personalization, Task (project management), Reuse

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange