Reproduce “CLOSER: Evidence-Aware Active Visual Verification for Embodied 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, Mingye Jiang, Yunze Hu, Zhejia Wen, Kangjie Zhou, Zongru Yang, Zhe Peng, Chang Liu
CLOSER lets object-goal agents defer uncertain detections, approach candidates, and reassess them from additional views. It combines a multi-view verifier and approach policy. Across eight systems and three simulated datasets, it improves success rate and path-weighted success in all 24 comparisons, averaging 3.6 and 2.0 percentage points. Physical-world evidence is limited to verification demonstrations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.