Reproduce “Weight-Space Programming of Policies and Their Verification via Model Checking” 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
Dennis Gross
The authors present a pipeline where an LLM writes neural-network weights for a decision policy and Storm model-checks its formal requirement. In small modeled environments, DeepSeek-V4-Pro policies for two transporter jobs averaged 59.52% dataset-state action accuracy yet achieved the optimal completion probability of 1.000 in all three runs. The evaluation covers only small environments and three runs per configuration.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.