Reproduce “CupCleaner: A Hybrid Data Cleaning Approach for Comment Updating” 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
Qingyuan Liang, Zeyu Sun, Qihao Zhu, Junhao Hu, Yifan Zhao, Yakun Zhang, Lu Zhang
CupCleaner cleans Java training and validation examples for code-comment updating using semantic relationships among comments and code, plus per-example loss and its variation during training. It filters examples that score poorly or remain difficult under its rules. On the 2021 dataset, CodeT5 reaches 45.33 exact match and 69.11 BLEU-4; this result is limited to the evaluated Java test data.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.