Reproduce “CovNav: Conformal Prediction-Guided Adaptive Inference for Object Goal Navigation” 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
Pengying Wu, Zhejia Wen, Kang Ding, Kangjie Zhou, Chunxuan Jiao, Wei Pan, Xiangli Nie, Chang Liu
CovNav lets a robot use a fast object-pair judgment unless conformal prediction marks it uncertain, then invokes the same local language model’s slower reasoning mode. Its coverage guarantee applies to object-pair labels under exchangeability, not navigation success. On full HM3D v1 and v2 test sets, success rates are 60.4% and 77.9%, versus ApexNav’s 59.6% and 76.2%.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.