Unlock: Weighted Conformal Prediction Under Covariate Shift

The extension of split conformal prediction that restores finite-sample coverage when the test distribution differs from the training distribution, at the cost of knowing or estimating the likelihood ratio between them.