nice article. You might be interested in conformal prediction, that jackknife+ is just one example of.

In particular Conformal Predictive Distributions don’t have any overhead in their inductive form and output the whole CDF for test point. I think CPDs implemented in new crepes library.

https://github.com/valeman/awesome-conformal-prediction

https://proceedings.mlr.press/v91/vovk18a.html

https://github.com/henrikbostrom/crepes/blob/main/crepes.ipynb