Reproduce “MarvisNav: Making Memory Visible on Route Choices for Zero-Shot Object 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
Jincheng Wang, Chi Pui Chan, Wei Zeng, Shuyang Zhang, Jianhao Jiao, Dimitrios Kanoulas
MarvisNav helps robots search unfamiliar buildings by attaching local exploration states to visible route choices, letting a vision-language model assess target relevance and unexplored space together. The system updates these states as the robot moves. On HM3D v0.2, it achieves 81.2% SR and 42.5% SPL, above separate visited markers in a test, though this combines state design and choice alignment.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.