開源AI代理框架與代理應用中測試實踐的實證研究

An Empirical Study of Testing Practices in Open Source AI Agent Frameworks and Agentic Applications

September 23, 2025

作者: Mohammed Mehedi Hasan, Hao Li, Emad Fallahzadeh, Gopi Krishnan Rajbahadur, Bram Adams, Ahmed E. Hassan

cs.AI

摘要

基于基础模型（FM）的人工智能代理正在多个领域迅速获得应用，但其固有的非确定性和不可复现性给测试和质量保证带来了挑战。尽管最近的基准测试提供了任务层面的评估，但开发者在开发过程中如何验证这些代理的内部正确性仍缺乏深入理解。 为填补这一空白，我们首次对人工智能代理生态系统中的测试实践进行了大规模实证研究，分析了39个开源代理框架和439个代理应用。我们识别出十种不同的测试模式，并发现诸如DeepEval等新颖的、专为代理设计的测试方法使用率极低（约1%），而传统的负面测试和成员测试等模式则被广泛采用以应对FM的不确定性。通过将这些模式映射到代理框架和代理应用的典型架构组件上，我们发现了一个根本性的测试努力倒置现象：确定性组件如资源工件（工具）和协调工件（工作流）占据了超过70%的测试努力，而基于FM的计划主体却仅获得不到5%的关注。尤为关键的是，触发组件（提示）几乎被忽视，仅出现在约1%的测试中。 我们的研究结果为基于FM的代理框架和代理应用提供了首个实证测试基线，揭示了在应对非确定性方面存在理性但不完全的适应。为解决这一问题，框架开发者应加强对新颖测试方法的支持，应用开发者需采纳提示回归测试，而研究者则应探索采用这些方法的障碍。强化这些实践对于构建更加健壮和可靠的人工智能代理至关重要。

English

Foundation model (FM)-based AI agents are rapidly gaining adoption across diverse domains, but their inherent non-determinism and non-reproducibility pose testing and quality assurance challenges. While recent benchmarks provide task-level evaluations, there is limited understanding of how developers verify the internal correctness of these agents during development. To address this gap, we conduct the first large-scale empirical study of testing practices in the AI agent ecosystem, analyzing 39 open-source agent frameworks and 439 agentic applications. We identify ten distinct testing patterns and find that novel, agent-specific methods like DeepEval are seldom used (around 1%), while traditional patterns like negative and membership testing are widely adapted to manage FM uncertainty. By mapping these patterns to canonical architectural components of agent frameworks and agentic applications, we uncover a fundamental inversion of testing effort: deterministic components like Resource Artifacts (tools) and Coordination Artifacts (workflows) consume over 70% of testing effort, while the FM-based Plan Body receives less than 5%. Crucially, this reveals a critical blind spot, as the Trigger component (prompts) remains neglected, appearing in around 1% of all tests. Our findings offer the first empirical testing baseline in FM-based agent frameworks and agentic applications, revealing a rational but incomplete adaptation to non-determinism. To address it, framework developers should improve support for novel testing methods, application developers must adopt prompt regression testing, and researchers should explore barriers to adoption. Strengthening these practices is vital for building more robust and dependable AI agents.