Reproduce “A Library for Representing Python Programs as Graphs for Machine Learning” 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
David Bieber, Kensen Shi, Petros Maniatis, Charles Sutton, Vincent Hellendoorn, Daniel Johnson, Daniel Tarlow
The python_graphs library builds Python program graphs linking syntax with execution paths and variable relationships. It provides control-flow graphs, data-flow analyses, and composite program graphs for machine-learning inputs or training targets. In Project CodeNet, it constructed control-flow graphs for 3,157,463 of 3,286,314 submissions (96.08%). These are construction results, not evidence of semantic correctness; analyses are best-effort.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.