Do Autonomous Agents Contribute Test Code? A Study of Tests in Agentic Pull Requests

TL;DR

This study analyzes how autonomous coding agents incorporate tests in pull requests using the AIDev-pop dataset. By labeling PRs that touch test files and computing lifecycle metrics (churn, turnaround, merge rate, and test-to-code churn), the authors show that test inclusion rises over time and varies across agents, with tests often touched early but revised later. Test-containing PRs are consistently larger and take longer to complete, though merge rates remain similar across agents; Codex exhibits particularly fast PR lifecycles, while Devin shows higher revision rates for test touches. The findings offer empirical grounding for understanding testing practices in agent-assisted software development and point to factors that shape test adoption in autonomous workflows.

Abstract

Testing is a critical practice for ensuring software correctness and long-term maintainability. As agentic coding tools increasingly submit pull requests (PRs), it becomes essential to understand how testing appears in these agent-driven workflows. Using the AIDev dataset, we present an empirical study of test inclusion in agentic pull requests. We examine how often tests are included, when they are introduced during the PR lifecycle and how test-containing PRs differ from non-test PRs in terms of size, turnaround time, and merge outcomes. Across agents, test-containing PRs are more common over time and tend to be larger and take longer to complete, while merge rates remain largely similar. We also observe variation across agents in both test adoption and the balance between test and production code within test PRs. Our findings provide a descriptive view of testing behavior in agentic pull requests and offer empirical grounding for future studies of autonomous software development.