Reproduce “DocPrism: Multi-lingual Detection of Incorrectness Inconsistencies between Code and Documentation” 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
Xiaomeng Xu, Zahin Wahab, Reid Holmes, Caroline Lemieux
DocPrism checks method documentation against code by asking a language model to classify inconsistencies, then filtering out under-promises, or omitted code behavior. In a labeled 52-pair Python ablation, its flag rate fell from 98% to 14% and F1 rose from 0.21 to 0.77 versus an instructed-filter variant. Across 1,991 pairs, it flagged 17% with 0.63 precision; recall was unmeasured.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.