navarog/cross-validated-ope ? reverse-engineered prompt

Build me a Python project that reproduces the experiments from a paper on cross validated off policy evaluation.

I want to be able to install the dependencies, run one command with a config file, and get the figures and saved results for the experiment. It should support replaying an existing saved run so I can replot results without rerunning all the heavy calculations, and let me tweak things like the plot title or which values are being compared. Use standard ML tools and PyTorch, and make the configs easy to edit so I can swap datasets, policies, estimators, and tuning methods without touching the code much.

Please include a simple example run that finishes quickly, plus the paper style experiments that save figures and intermediate data into organized results folders. If you need to check current library docs or best practices, go ahead and look them up online.