Reproduce “Amortized Score-Hamiltonian Policy Iteration: A Grid-Free Scheme for Relaxed Stochastic Control Problems” 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
Qi Feng, Gu Wang
The method replaces state-specific Langevin samplers with a shared conditional sampler and value-function critic trained across sampled states. It projects score-mismatch transport onto motions the sampler can produce, avoiding a state grid while assuming a known model. In a 16-control test, mean total-variation distance across one-dimensional marginals fell from 0.597 to 0.128; this does not establish joint-policy accuracy or performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.