Reproduce “LLMs as MDP Designers: A Dependency-Aware Agentic Framework for Automated Robot Policy Generation” 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
Hong Je-Gal, Jinwoo Kim, Chanbin Yi, Hyun-Suk Lee
The authors present ARPG, which uses language-model agents to formulate a robot’s Markov decision process before simulator implementation; its resulting policy runs without language models at deployment. Across five Isaac Lab tasks, ARPG has the highest policy-generation success rate on each. On Reach, modeling, implementation, and policy success rates are 0.9 each, versus Codex’s 0.4, 0.3, and 0.2; evidence is simulated.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.