Reproduce “Categorical Internalisation of Environmental Groupoids for Generalisable POMDP Solving” 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
Ben Opperman, Eduardo Alonso, Esther Mondragón
The authors present groupoid-based Q-learning that shares experience across locally equivalent states. It maps states to canonical representatives, records transformations back to encountered states, and propagates action-value updates. On 20×20 RockSample, structural canonicalization stabilizes near reward 25 within about 500 episodes, versus about 23 for automatic canonicalization; these are benchmark reward results, not runtime measurements or evidence of real-world performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.