Reproduce “Advantage-Driven Explicit Memory for Social 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
Yeonsoo Park, Mattia Racca, Guillaume Bono, Steeven Janny, Gianluca Monaci, Tomi Silander, Christian Wolf
The authors present an online-memory policy for social navigation that retrieves recurrent hidden states linked to earlier actions and outcomes. It stores rare high-advantage events, retrieves the top five similar situations, and adds test-time experience without gradient updates. In simulated out-of-distribution PointGoal navigation, success rate, the percentage of successful episodes, rose from 56.40% to 60.00%, but real-robot safety remains untested.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.