Compressed Active Subspaces for Scalable Bayesian Inference

A new method for scalable Bayesian inference that could improve uncertainty quantification in large models.

Auto-scored: Technical math/statistics paper with no direct AI risk or mitigation implications.