Reproduce “DEPLOY-RL: Active Boundary Discovery and Conservative Certification for Deployable Reinforcement Learning in Safety-Critical Continuous Processes” 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
Yeojin Jang, Minu Baek, Gihun Gil, Minsung Jung, Beomdo Park, Woohyeon Kwon, Hyeonseok Jang, Junseong Park and 3 more
DEPLOY-RL is a post-training certification layer that decides whether to deploy a reinforcement-learning controller, reject it, or defer to PID/MPC fallback. It uses contract-weighted active simulator sampling near decision-relevant boundaries, then conformal calibration to make the decision. Across papermaking and Tennessee Eastman, it reports 4.4% false-go and 88.6% coverage versus 8.6% and 79.8% for Deep Ensemble under simulator contracts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.