This paper constructs a large-scale dataset of agent applications, tools, and tests, and manually label 2,572 test methods from 240 modules, and derives a taxonomy of 23 testing patterns across test fixtures, data, objectives, and assertions, and characterize tests by level.
Abstract
Agents built on large language models (LLMs) are increasingly used to build applications that perform complex, multi-step tasks involving reasoning, tool use, and interaction with external environments. Despite rapid progress in benchmarking LLM-based agents, very few studies have attempted to understand how such systems are tested in practice. In particular, testing levels, objectives, data patterns, test complexity, and validation strategies for agent applications remain underexplored. In this paper, we present an empirical study of testing practices in LLM-based agent applications using a large corpus of mined open-source projects. We construct a large-scale dataset of agent applications, tools, and tests, and manually label 2,572 test methods from 240 modules. From this analysis, we derive a taxonomy of 23 testing patterns across test fixtures, data, objectives, and assertions, and characterize tests by level (unit, module, integration). We complement this with structured interviews of 10 senior industry practitioners building agentic systems. Our results show that testing of LLM-based agent applications is dominated by narrowly scoped unit tests, with limited coverage of complex interactions, realistic scenarios, and non-functional requirements. Tests frequently rely on simplistic inputs, heavy mocking, and shallow validation, and agent-related tests exhibit low structural complexity. While industry practice places greater emphasis on non-functional testing than open-source projects, both reveal common gaps, including the lack of formal testing foundations, unclear test objectives, and challenges in generating high-quality test data. Based on these findings, we outline research directions toward more systematic and rigorous testing of agent applications, including foundations for agent testability, formalized test objectives, and fault-based testing techniques.
Recently, the emergence of Large Language Models (LLMs) has spurred a surge of research into automated unit test generation, yielding impressive performance and reducing manual effort. However, existing LLM-based approaches still suffer from two major limitations: (1) they follow rigid, procedural workflows that underutilize the autonomous reasoning potential of LLMs, making it difficult to dynamically adapt testing strategies based on real-time feedback; and (2) they rely on rule-based context extraction that is not tailored to test generation, failing to capture fine-grained code dependencies and test-specific knowledge required for deriving test requirements. In this paper, we propose TestAgent, an LLM-based test generation approach that addresses the above limitations by emulating human testing practices via a multi-agent collaboration mechanism. Particularly, TestAgent designs three specialized agents, namely a requirement planner, a test generator, and a test reviewer, to simulate how developers understand, construct, and validate unit tests. To unleash the autonomous capabilities of LLMs, we equip TestAgent with a set of tool APIs that can be invoked dynamically in an on-demand and adaptive manner. To further support repository-level reasoning, TestAgent constructs a test-specialized knowledge graph via static analysis, which captures code entities and their dependencies across the project and persistently stores testing artifacts (e.g., test reports and failure analyses) produced during generation. Experimental results show that TestAgent achieves 97.46% execution rate, 92.34% line coverage, 90.24% branch coverage, and 83.69% mutation score on six Java projects, outperforming LLM-based baselines across all metrics and achieving substantially higher mutation scores than search-based tools.
Quanjun Zhang, Ye Shang, Siqi Gu et al.· 0 citations
LLM-based agents are increasingly used for coding tasks, where they have outperformed many classical approaches and scaled to repository-level tasks, such as test generation. However, when directly prompted to generate tests, these agents can fail to reason about the code and its underlying contracts, thereby missing edge cases and behavioral boundaries that affect test quality. To address this limitation, we propose Spec-Driven Test Generation, where we instruct an agent to first reason about -- and explicitly document -- code pre-conditions, post-conditions, and undefined behaviors. This intermediate semi-formal specification acts as a cognitive scaffold to guide subsequent test generation. Our evaluation on production bugs from Google shows that the spec-driven agent can deliver a 9.8 percentage points ($p = 0.0352$) improvement in bug detection rate and a 2.5 percentage point ($p = 0.0034$) improvement in branch coverage, compared to a traditional test generation agent baseline. Using LLM-as-a-Judge, we further show that test suites generated by the spec-driven agent are superior to the baseline and human-authored tests in 77.8% and 56.7% of the cases, respectively, and demonstrated improvements on following best practices, readability, and edge-case coverage.
Michele Tufano, James McClure, José Cambronero et al.· 0 citations
A large-scale empirical study of mainstream open-source agent frameworks from an engineering perspective, providing empirical evidence linking framework design choices to engineering risks and highlighting the need for stronger guidance and support in agent framework development.
Yibo Zhai, Junjun Si, Yan Wang et al.· SIGSOFT FSE Companion· 0 citations
TestAgent, a multi-agent tool implemented as a VS Code extension that automates the generation of high-quality unit tests for Java projects using repository-level Code Knowledge Graphs, demonstrates its practical utility for regression testing and bug discovery.
Ye Shang, Quanjun Zhang, Zheng Zhan et al.· SIGSOFT FSE Companion· 0 citations
Language agents, i.e., LLM agents, progress rapidly and are increasingly deployed in production environments. This trend underscores the urgent need for rigorous and realistic evaluations. However, most existing benchmarks evaluate agents in simplified, idealized settings. They typically rely on pre-packaged tool interfaces, overlook critical steps, and assume inputs are clean and fully specified. Consequently, they understate the difficulty of real deployments, where uncertainty and noise are ubiquitous and agents must proactively explore the environment to uncover new tools. To bridge this gap, we present AgentGym2, a new evaluation framework with task instances grounded in real-world end-to-end working demands. Beyond reasoning and planning, it measures agents'ability to execute end-to-end procedures, discover tools via exploration, compose tools for unseen tasks, and remain robust to noisy and underspecified information. Experiments on 15 proprietary and open-source models show that even SOTA systems like Gemini and GPT-5 struggle on AgentGym2, revealing a substantial gap between the capability of current agents and the demands of real-world applications.
Zhiheng Xi, Dingwen Yang, Jiaqi Liu et al.· Annual Meeting of the Associ...· 1 citation
Research on integrating specialized, language-model-based agents for automated test-case generation remains limited, and the detection criteria under which such systems are evaluated have not been standardized, which makes reported rates difficult to compare across studies. This paper presents a multi-agent testing framework in which requirement analysis, test-case generation, sandboxed execution, and defect detection are delegated to four distinct agents, and in which detection rests on a composite rule combining the execution signal with the semantic judgment of a dedicated diagnostic agent. The framework is evaluated on the QuixBugs dataset of forty Python programs under three criteria of increasing strictness. A two-run differential protocol against the reference implementation, recommended here as the primary indicator, yields a detection rate of 55.0%, against 90.0% under the unconditional criterion conventional in prior work and 5.0% under a strict criterion requiring the suite to pass entirely on correct code. The thirty-five-point gap is traced to language-model test artifacts, of which hallucinated oracle values and signature mismatches account for the majority. An ablation with a same-model single-agent baseline attributes the framework’s advantage to role decomposition rather than to the diagnostic agent specifically, and the distribution of detection across the fourteen defect classes defined by the benchmark is reported descriptively.
Yuxuan Li· Transactions on Computing Sc...· 0 citations