Reproduce “Skeleton-Guided-Translation: A Benchmarking Framework for Code Repository Translation with Fine-Grained Quality Evaluation” 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
Xing Zhang, Jiaheng Wen, Fangkai Yang, Pu Zhao, Yu Kang, Junhao Wang, Maoquan Wang, Yufan Huang and 6 more
The authors present Java-to-C# repository translation that preserves a compilable skeleton and tests relevant translated code within it. They define build success as the share of unit tests that compile and unit-test success as the share of compiled tests that pass. In a GPT-4o ablation, skeletons raise rates from 3.3% to 58.17% and 17.97%, respectively, in one 13-task setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.