Reproduce “ForagerSim: Evaluating Embodied Map-Based Reasoning in Large-Scale Dynamic Worlds” 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
Tomáš Musil, Martin Kotík, Michal Vavrečka, Karla Štěpánová, Sebastian Scherer, Martin Saska
ForagerSim provides a configurable simulator for testing whether embodied agents need map-based reasoning. Its comparisons use a simplified 2D sensor, contrasting unordered object search with trips that require returning to an unmarked home location. In a four-arm maze, DreamerV3 averages 272 ± 131 return, versus 1,133 ± 119 for one human player; this suggests a bottleneck, not its sole cause.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.