Reproduce “RIDAR: Safeguarding Research Integrity in Autonomous Research Agents via Automated Detection and Runtime Intervention” 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
Jiening Siow, Amy Xin, Junjie Wang, Ziyu Tang, Zijun Yao, Yunjia Qi, Fanjin Zhang, Lei Hou and 2 more
RIDAR audits research agent trajectories for integrity violations and can monitor running tasks. It uses a 14-type taxonomy and checks proposed findings against task rules and evidence. In evaluation, its post-hoc detector achieved 85.2% instance-level F1 on 60 expert-annotated trajectories. Runtime intervention was associated with 36.8% fewer accepted findings remaining at completion in matched reruns, within the tested setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.