Reproduce “Challenging Bug Prediction and Repair Models with Synthetic Bugs” 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
Ali Reza Ibrahimzada, Yang Chen, Ryan Rong, Reyhaneh Jabbarvand
BUGFARM generates synthetic Java bugs by prompting an LLM to alter multiple low-attention statements, rejecting invalid changes. In an 11-statement example, it changes an initialization and return to null. Tested predictors found its bugs harder to detect; FitRepair patched 28% of one sampled set. These synthetic bugs complement real-world bugs and do not establish difficulty for classic detectors or all repair tools.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.