Reproduce “LERR-Nav: Learning to Coordinate Exploration, Recovery, and Reminiscing for 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

Jingzhi Huang, Junkai Huang, Yuhang Peng, Keyu Mu, Hailong Huang, Haoyang Yang, Renbo Luo, Lujia Wang and 1 more

LERR-Nav coordinates value-function-guided exploration, LLM-guided exploration, recovery, and reminiscing through a high-level policy that selects a strategy after each execution ends, using summaries and visual records of activity. In HM3D episodes where the robot and target start on different floors, it achieves 62.5% success and 27.3% SPL, versus ASCENT’s 33.3% and 14.9%; this full-system comparison does not isolate memory’s effect.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.