Reproduce “RECTIFY: An Interactive Workbench for Post-Evaluation RAG Diagnosis, Repair, and Verification” 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
Keerthana Murugaraj, Salima Lamsiyah, Martin Theobald
RECTIFY turns evaluated retrieval-augmented generation failures into named repair slices and editable repair cards. It filters cases, routes them through four failure families and 23 slices, and lets developers approve, reject, or optionally test proposals. In a 100-question synthetic benchmark, estimated repair decisions fell 97% versus raw scanning, but results do not establish faster debugging or guaranteed fixes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.