Reproduce “Generalizing Causal Decision-Making: Markov Equivalence, Transportability and Counterfactuals” 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
Min Woo Park, Sanghack Lee
The authors develop structural causal bandit methods that use causal knowledge to remove redundant or suboptimal actions before learning rewards. They address uncertain graphs, cross-environment information, and realizable counterfactual experiments. On a synthetic model, a reduced set retains 54 actions and yields mean regret 31.0 versus 4,929.9 for brute force; this result uses an oracle graph and does not establish performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.