interesting article, the main issue with quantile regression is that of course it doesn’t provide unbiased prediction and has no calibration guarantees. So the results produced by using quantile regression won’t be useful in high stakes applications and will lead to risky decisions.

https://valeman.medium.com/how-to-predict-quantiles-in-a-more-intelligent-way-or-bye-bye-quantile-regression-hello-24a65e4c50f

These days it is possible to produce the whole CDF for each test object without using parametric methods such as quantile regression and the conformal prediction methods have in built validity guarantees and do not use parametric assumptions

https://valeman.medium.com/how-to-predict-full-probability-distribution-using-machine-learning-conformal-predictive-f8f4d805e420

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

Quantile regression is very old tool (1978) and the author of quantile regression himself has recently discovered conformal regression.

https://www.linkedin.com/feed/update/urn:li:activity:7026650490859880449?updateEntityUrn=urn%3Ali%3Afs_feedUpdate%3A%28V2%2Curn%3Ali%3Aactivity%3A7026650490859880449%29