Reproduce “SWE-Nexus: Automated Composition of Software Engineering Tasks from Test Execution” 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
Songwei Yu, Ziqi Miao, Jingxuan Xu, Junhao Fan, Haoyang Huang, Yuan Xiong, Zihang Xu, Weihao Li and 5 more
SWE-Nexus constructs repository software-engineering tasks from production function bodies exercised by selected tests, then combines compatible scopes and rechecks every constituent test under the combined mask. Its benchmark contains 600 tasks across six languages, including 484 composite tasks. Under Claude Code, Claude Opus 4.8 resolves 37.8% at one attempt, while validation establishes relevance only to selected tests.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.