The paper presents XUAT-Copilot, a multi-agent collaborative system developed to automate the User Acceptance Testing (UAT) process of WeChat Pay, a leading mobile payment app in China. The system aims to increase efficiency in generating test scripts, a labor-intensive stage in the current system. XUAT-Copilot leverages Large Language Models (LLMs) to mimic human intelligence and decision-making capabilities. The system includes three LLM-based agents for action planning, state checking, and parameter selecting, and two modules for state sensing and case rewriting. The system has been launched in the formal testing environment of WeChat Pay, demonstrating effectiveness and accuracy in its performance.
Publication date: 8 Jan 2024
Project Page: https://doi.org/XXXXXXX.XXXXXXX
Paper: https://arxiv.org/pdf/2401.02705
