Loss-Driven Bayesian Active Learning
arXiv:2604.11995v1 Announce Type: new
Abstract: The central goal of active learning is to gather data that maximises downstream predictive performance, but popular approaches have limited flexibility in customising this data acquisition to different d…