Reproduce “From Tokens to Files: Multi-Granularity Detection and Localization of AI-Involved Code” 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
Yue Xu, Xiyang Wang, Hui Liu, Bolei He
The system detects and localizes AI-involved code within files. It compares each token’s likelihood under AI-adapted and human-adapted versions of one frozen code model, then uses context-aware heads to estimate AI involvement by line, block, and file. On six synthetic test regimes, it achieves 72.4% average macro F1; this internal result does not establish performance on AI-assisted code.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.