Reproduce “SWE-at-Scale: Scalable Issue Synthesis from Historical Bug Fixes” 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

Silin Chen, Zihan Zhang, Xiaodong Gu, Jiale Huang, Yuling Shi, Han Li, Zhao Shan, Chengcheng Wan and 2 more

The authors introduce SWE-at-Scale, which learns to generate repository-level bug candidates by reversing real fixes into training examples and applying transformations to correct code. Its pattern-based variant describes bug logic in natural language before an executor modifies code. On matched files, bug-type entropy rises from 0.583 to 2.063 bits through 200 samples; this Python study does not establish cross-language generalization.

Summary by alphaXiv.

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