Reproduce “Unit Test Update through LLM-Driven Context Collection and Error-Type-Aware Refinement” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yuanhe Zhang, Zhiquan Yang, Shengyi Pan, Zhongxin Liu
TESTUPDATER updates Java unit tests after production-code changes by collecting relevant method definitions and existing test-class fields, then using execution errors to refine generated tests. On 195 UPDATES4J examples, DeepSeek-V3 achieves 94.4% compilation and 86.7% test pass rates, versus SYNTER’s 78.5% and 66.7%. The benchmark is mostly repairs; enhancement-only results are unreported.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.