From Feedback to Failure: Automated Android Performance Issue Reproduction

2025-08-14

AI Summary

Performance issues in mobile applications are notoriously difficult to reproduce and diagnose within development environments. To address this challenge, we propose an automated performance-problem reproduction method grounded in user reviews. First, large language model–based prompt engineering is employed to infer missing operational context and semantic information from sparse or ambiguous review texts. Second, an execution agent translates the enriched textual descriptions into executable UI interaction sequences to reliably reproduce the reported issues on physical devices. Third, multi-source validation integrates Android system logs, GUI change detection, and multidimensional resource monitoring (e.g., CPU, memory, battery) to confirm reproduction fidelity. This work represents the first approach to synergistically combine user-review analysis with prompt engineering specifically for performance-issue reproduction. Evaluated on a manually annotated benchmark dataset, our method achieves a 70% reproduction rate—substantially improving both the efficiency and interpretability of performance defect localization.

Technology Category

• Search and Optimization: Evaluation and Analysis
• Natural Language Processing: Prompt Engineering / Prompting
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success

Abstract

Mobile application performance is a vital factor for user experience. Yet, performance issues are notoriously difficult to detect within development environments, where their manifestations are often less conspicuous and diagnosis proves more challenging. To address this limitation, we propose RevPerf, an advanced performance issue reproduction tool that leverages app reviews from Google Play to acquire pertinent information. RevPerf employs relevant reviews and prompt engineering to enrich the original review with performance issue details. An execution agent is then employed to generate and execute commands to reproduce the issue. After executing all necessary steps, the system incorporates multifaceted detection methods to identify performance issues by monitoring Android logs, GUI changes, and system resource utilization during the reproduction process. Experimental results demonstrate that our proposed framework achieves a 70% success rate in reproducing performance issues on the dataset we constructed and manually validated.

Problem

• Detect Android app performance issues from user reviews
• Reproduce performance issues using automated execution and monitoring
• Improve issue diagnosis with multifaceted detection methods

Innovation

• Leverages app reviews for performance issue detection
• Uses prompt engineering to enrich review details
• Monitors logs, GUI, and resources for issue identification