Reproduce “What You See Is What You Get: Attention-based Self-guided Automatic Unit Test Generation” 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

Xin Yin, Chao Ni, Xiaodan Xu, Xiaohu Yang

AUGER, an attention-based self-guided system, enhances software defect detection and remediation by integrating defect prediction with large language model-guided unit test generation. It improves developer confidence and efficiently triggers errors by focusing test generation on identified defective code segments, achieving higher error-triggering rates than existing methods.

Key insights

  • Combining adversarial and contrastive learning significantly boosts the robustness and generalization of defect detection models, enhancing their ability to identify real-world defects across different datasets.
  • Directing the attention of LLMs to specific defective code sections through targeted attention head modification substantially improves the efficiency and effectiveness of generating error-triggering unit tests.
  • An integrated 'What You See Is What You Get' system, where defect detection informs and guides test generation, provides concrete validation for detected defects, thereby increasing developer confidence and trust in AI-based tools.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.